A Multimodal Multidimensional (4D) Map of the Mouse Brain
A Multimodal Multidimensional (4D) Map of the Mouse Brain
Project Overview
This is a project to develop a detailed multidimensional digital atlas of the mouse nervous system. It will, for the first time, enable a unified framework for representing brain maps and gene expression maps. It will provide the ability to chart the anatomy of gene expression in the brain of adult and developing mice. It will establish the linkage between genotype and phenotype. It will enable the comparison of gene expression maps from different animals, laboratories and strains.
Our goals are to collect data from multiple modalities (MRI, PET, blockface imaging and histology), reconstruct these data, place them in a defined coordinate system, delineate and describe the anatomy and develop the appropriate informatics tools to interact with these data. We will do this for adults and developing animals at various maturational stages. We will use the C57BL/6J mouse strain. The product of these efforts will be a series of comprehensive digital atlases describing the probabilistic neuroanatomy of the mouse, a set of tools to import images of gene expression into the atlas and tools for interacting with the atlas statistically and visually in 4D.
This is an enormous undertaking. We are cognizant of the technical, scientific and labor intensive difficulties associated with this project. However, we are experienced in all the appropriate neuroscientific and neuroinformatics disciplines. We have performed preliminary experiments for every aspect of the research plan. We have previously developed atlases of other species along with useable and distributable software that demonstrates our ability to deliver mature products to the community.
The research plan includes three performance sites; UCLA, USC and CALTECH. Each was chosen because of the participant’s ability to provide significant expertise in each of the requisite elements of this proposal. Each complements the other resulting in a powerful team of investigators with unique and relevant resources and experience necessary for this project. We have already established productive working relationships.
Our plans also call for the use of workshops to help coordinate the research and development between our efforts and those of other grantees as well as the gene mapping community at large.
The structure of this proposal is of an integrated and unified project. In spite of the participation from multiple institutions, the research plan describes the work without identifying the responsible investigator. The budget and its justification provide details regarding the allocation of funds and the specific activities supported.
Specific Aims
The overall goals of this project are to develop and implement a probabilistic atlas of the adult and developing C57BL/6J mouse. These atlases will describe the mouse nervous system in detail within a well defined coordinate system. Since the atlas will be based on multiple subjects, it will describe the morphological variability within this strain as well as its development at maturational stages from in utero to adult. The data will be comprised of in vivo and post-mortem imagery. Tools for visualizing, measuring, spatially normalizing (deformation correction) and mapping gene expression data will be created and validated.
In addition to these design driven goals, we will test several hypotheses using the aforementioned atlases and tools. The product of this effort will be an atlas system for the C57BL/6J mouse, tools for using it and evidence that it enables the linkage between brain mapping and gene mapping.
Goals
1. To develop and implement an anatomic framework to map gene expression in the brain. This framework will be comprised of novel imaging data including μMRI, μPET, blockface imaging and histology.
2. To create a set of tools for colocalizing data from different markers, animals and laboratories.
3. To collect multimodality data describing the brain of the C57BL/6J mouse strain.
4. Dissemination of the map (and requisite interactive tools) enabling – output (ability for others to use information/data) and input (ability to incorporate data from other sources).
Hypotheses
1. There is a relationship between gene expression and morphology.
2. Patterns of gene expression co-vary with morphological changes during development.
3. Anatomy from histological delineations accurately represents in vivo morphology.
4. Within strain morphometric variability will be less than between strains.
The experiments necessary to test these hypotheses will utilize the atlases and associated tools along with gene expression maps (GEMS) collected as part of this study. In this way we can begin to test the utility and validity of our multimodality, multidimensional mouse atlas.
Background and Significance
Mapping the genome
The mouse is a living encyclopedia of known gene functions and a repository for the unknown gene functions that produce a developing, metabolizing, reproducing mammal. Consisting of up to one hundred thousand genes, the task of sequencing and cataloging the entire mouse genome is one of enormous scope and complexity. However, as mouse genome project advances, there has been a great demand for the structural, functional, and anatomic correlates of gene expression. Genetic maps have localized genes to specific sites on chromosomes, but their pattern of expression has only begun to be touched upon. Even so, there exists no coherent framework for the cataloging and comparison of gene expression. What is required is a Spatial and Temporal Atlas of Gene Expression (STAGE) to coordinate the collection and analysis of the enormous amount of data generated by the genome project.
Mice provide many advantages as a model system for mammalian genetics. Short generation time, large litters, and relatively low cost of care, make it a pragmatic choice. The existence of inbred strains of mice, where every individual (of the same sex) is genetically identical, enhances the information content of STAGEs, allowing for the collection of data from multiple individuals, secure in the knowledge that there will be no variation due to genetic factors, something not readily accomplished or impossible in other (outbred) species. The comparison across different strains of mice (i.e. C57BL/6 and Spret/Ei) will allow for analysis of discrete genetic differences which may already be cataloged (Lyon, 1996). In addition, the generation of transgenic (gain-of-function) and targeted knock-out (loss-of-function) mice represent a new kind of mapping directed toward the elucidation of gene and genome function by the time-honored genetic approach of comparing mutant and normal phenotypes. Combining genetic remodeling, genetic maps, banks of genes and emerging methods for assessing gene expression on a whole tissue level, the true potential of the mouse as a mammalian model is beginning to emerge.
Gene expression can be studied in a number of ways. Measurements of mRNA within a tissue can be done by traditional electrophoretic techniques such as Northern blotting and RT-PCR. Both are technically simple, but limited by the crudity of harvesting tissue manually. Cytochemical methods such as in situ hybridization can also measure the expression of a specific mRNA, but in turn are limited by their qualitative nature. Furthermore, gene expression can also be measured by the level of protein expressed. These measurements can be made either electrophoretically by Western blot or by cytochemical methods such as immunohistochemistry. In both cases protein levels are measured by the binding of an antibody specific for the protein itself, making this perhaps the best indicator of gene expression. (Should microPET be mentioned here?)YES AND POINT TO PRELIM RESULTS
Atlases and Maps
Atlases of normal mouse development have immense pedagogical value and provide researchers studying normal, mutant, and transgenic mice a standard against which specific examples may be compared and contrasted. Standard methods of atlas construction typically involve sacrificing, fixing, sectioning, staining, then recording photomicrographs of individual sections. Photographic plates are the raw material of most atlases that contain three additional critical elements: 1) annotation in the form of graphical reconstructions highlighting important detail; 2) nomenclature in the form of descriptions and names of discrete structures; and 3) imposition of a 3D coordinate system so that anatomy can be referred to using a standardized atlas. Atlases of this type for the mouse have been presented by Rugh (1990), Theiler (1989) and Kaufman (1992). The advent of powerful inexpensive computers coupled with the ability to conveniently transport large amounts of data (via CD-ROM or over the Internet) are bringing about changes in the way atlases are constructed and in the ways they can be used. When in book form, the intrinsically 3D animal must be viewed as a series of 2D sections. Moreover, the orientations available to the viewer are limited to samples of standard planes of section (e.g. sagittal, coronal, axial). These restrictions make it difficult to follow complex 3D structures and hinder comparison of one's own 'oblique' sections with the 'perpendicular' sections found in the atlases. Digital atlases have the potential to obviate both of these vexing problems (Williams & Doyle, 1996; Kaufman, et al., 1997; Gibaud et al., 1997 and Toga et al., 1995). With the section data reconstructed into three dimensions, highlighting complex structures and computationally sectioning at arbitrary angles becomes possible. Quantitative morphological measurements (volumes, distances, angles) can be accomplished and maps can be generated that amalgamate data from various experimental techniques. Temporal and spatial gene and protein expression patterns, axonal trajectories, patterns of vasculature, and specific neuronal responses to stimuli can all be combined to obtain a canonical organism or system. Such a data set could potentially embody all quantitative information known about the animal in a concise framework.
Informatics
Motivated by such benefits, several efforts are underway to generate digital atlases. There is at least one commercially available CD-ROM rat atlas (Paxinos and Watson, 1991) and other less ambitious CD-ROM undertakings (Ghosh, et al., 1994; Smith, et al., 1996). A number of World Wide Web sites present a variety of two dimensional data (rodents) and some aim towards being three dimensional atlases (www_atlases; Toga et al., 1995). Based upon the atlas of the developing mouse (Kaufman, 1992) the Edinburgh group has embarked on a significant effort to create a database to house gene expression (Ringwald et al., 1994). In our own laboratories, the ICBM effort at mapping the human brain, is based upon a digital 3D representation of a population’s anatomy. The spatial normalization, warping, morphometrics, visualization, databasing and related informatics efforts included in brain mapping have made enormous progress in the last few years (Koslow & Huerta, 1998; Toga & Mazziotta, 1996,1999; Toga, 1999). Many of these advancements have direct relevance to the present project and specific examples are provided below.
|Name | |Reference | |Features |
|Muritech | | | | |
|UT Memphis | | | | |
|Edinburgh | | | | |
|Chemoarchitectonic Atlas of the | |Jacobowitz & Abbott, 1997 | |Calretinin, calbindin, serotonin, tyrosine |
|Developing Mouse Brain | | | |hydroxylase, AChE, 336 dpi/20 mm |
|Atlas of the Prenatal Mouse Brain | |Shambra, Lauder & Silver, 1992 | |H & E; 10mm |
Technology
High resolution MR. The use of magnetic resonance imaging has revolutionized the noninvasive investigation of neuroanatomy and function, and is an integral component of digital atlasing. Recent reports of the observation of subtle intracortical structure such as the stria of Gennari (Clark et al., 1992), the histological confirmation of the normal MR distribution of the corticospinal tract (Yagishita et al, 1994) and the ability to selectively image myelin (MacKay et al, 1994) support efforts to apply MR techniques to the imaging of anatomy. The combination of higher field magnets and post processing techniques for image enhancement (Kui Ying et al, 1996; Holmes et al, 1997) has also recently revealed structures as fine as thalamic nuclei, the origin of the thalamocortical tracts, and there is evidence that we can discriminate in vivo cortical architectonic regions. These technical innovations permit the macroscopic observation of fascicles through the in vivo and cranially-intact post mortem brain. Even greater strides can be taken by using extremely high field instruments in the smaller primate species, using microscopic MRI.
Microscopic MRI. The notion of using MRI at microscopic resolutions arose early in the development of this technique (Lauterbur 1973). The spatial resolution in biological samples is typically limited by line-width broadening (T2 effects), diffusion, signal-to-noise ratio (S/N), factors whose physical limits have been discussed in detail by Callaghan (Callaghan 1991) and others (House 1984; Cho et al. 1988; Kuhn 1990; Blumich and Kuhn 1992; Zhou and Lauterbur 1992). Estimates of the theoretical limits of resolution in the MR image range from 2 to 0.5 micron (Cho et al. 1988; Kuhn 1990; Callaghan 1991). By judicious choice of experimental conditions (e.g. bandwidth and gradient strength), deterioration in resolution due to the combination of these effects can be compensated to a degree, resulting in a practical spatial resolution, currently limited by the amount of time available to acquire the image. The quality and usable resolution of these MRI images can be enhanced in several ways, such as: increasing the main magnetic field strength to 12T ; optimizing the radio frequency (RF) coil for small samples; employing 3D volume imaging rather than slice imaging; and using fast imaging pulse sequences (e.g. DEFT, FLASH, EPI). Indeed, several groups have achieved spatial resolutions of 10 micron or less (Blumich and Kuhn 1992), and MRI of rodent eyes and xenographs correlates well with subsequent histological examinations of the same tissues. Aguayo and coworkers (1987) resolved cell clusters and structures as small as basement membranes.
Features and Benefits
The creation of a comprehensive framework capable of encompassing diverse information about the mouse holds tremendous promise for integrating the genotype and phenotype of this animal. Genetic information is expressed in complex and ever-changing patterns throughout the development of the animal. A comprehensive description of these patterns and how they relate to the emerging morphology is crucial to our understanding of the interactions that underlie the processes of development, normal structure and function, disease and evolution.
Studies of gene expression are rapidly producing a vast amount of information relating to these complex patterns. It is currently impossible to adequately compare results from different animals, investigators and laboratories. It is also difficult to make comparisons between the expression of different genes in order to assess the possibility of complex networks of genetic interaction. These problems cannot be addressed by conventional means of publication, but require the development of an electronic database, together with tools for cross-modality correlation. Moreover, text descriptions of gene expression are of limited value due the spatial complexity of the patterns and partly because domains of gene expression do not necessarily correspond to named anatomical structures. The proposed multimodality, multidimensional atlas will address these limitations.
Preliminary Results and Expertise
Instrumentation
MRI. In this section we outline preliminary results obtained using our current 11.7T vertical bore MRI instrument. We focus on four topics demonstrating the feasibility of using µMRI as an anatomical survey methodology for obtaining 3D high resolution data of intact mice in vivo:
• µMRI of fixed mouse embryos at specific gestational ages
• In vivo in utero MRI of a day 12 pregnant mouse
• RF & gradient coil optimization
• Diffusion tensor MR imaging of a transgenic MS mouse model system
We begin with a discussion of 3D MR imaging of fixed mouse specimens which demonstrates the excellent image quality, resolution, and contrast that can be obtained on immobile samples in a several hour experiment (Jacobs et al., 1999). Preliminary diffusion tensor imaging of a transgenic mouse model for multiple sclerosis illustrates the wealth of information inherent in this imaging modality and its ability to delineate nerve tracts and abnormalities in the mouse spinal cord. We have also been able to collect multislice images of an anesthesized pregnant mouse indicating that it is feasible to obtain reasonable MR images in a 15 minute experiment with a live mouse. Finally, we comment on some of our experience building MRI probes (specifically RF & gradient coils) for our 11.7T systems which affirms that customizing this portion of the hardware can provide significant benefit at relatively modest cost in time and effort.
[pic]
Figure 1. Volume rendering of MR images of fixed mouse embryos. Gestation days are noted. The Day 6.5 image (left panel) has been sliced in half to show internal details. The Day 14 embryo is rendered semi-transparent to show internal anatomy. In vivo, in utero MR images of Day 12 mouse embryos. The mother and RF coil are shown to the upper right. A single slice showing several adjacent embryos is shown to the upper left. Lower panels show serial longitudinal slices through a single live embryo. (right panel). A higher resolution version of this figure appears in the Appendix.
MRI of Fixed Mouse Embryos at Different Gestational Days. Figure 1 shows renderings of 3D MRI data of fixed mouse embryos. Each specimen was excised and fixed at the gestation day noted. A single sagittal slice from the 6.5day and the 11.5day specimen are displayed, while volume renderings of the other datasets are shown. The datasets were all acquired with a 3D spin echo protocol with TR=880~1000ms. Excellent contrast is seen in the developing nervous system (see slice data day 11.5 and semitransparent day14). In this study, good contrast was obtained with short TE's (~30ms) for the young samples, but longer TE's (~100ms) were necessary to obtain good contrast in the older samples. Spatial resolution is 15microns in the 6.5-10.5day samples, 30microns for the 11-13day samples, and 60microns for the older samples. Imaging time ranged from 3-12hours in most cases, but was 24hours for the larger 16day specimen. Ventricles are always easily distinguished from the surrounding tissue. The developing heart is seen as early as day 8.5 (image not shown) and the rudiments of the skeletal system are apparent in the day14 image. The RF coil used was matched to the sample size to ensure maximum filling factor.
THESE IMAGES DEMONSTRATE THAT:
• spatial resolution ~50microns (isotropic??????) affords detection of many internal features;
• contrast changes with gestation day and may be enhanced by tuning experimental parameters;
• good signal to noise can be obtained in a reasonable time, even for the 30micron resolution images;
• imaging times are long, but not excessive for large 3D datasets.
MRI OF AGED MICROCEBUS. WE COMPARED BRAINS OF YOUNG AND AGED MICROCEBUS USING SEVERAL 2D AND 3D MRI ALGORITHMS AT 11.7T AND STANDARD HISTOLOGICAL EXAMINATION OF THE SAME TISSUE. THE SMALL SIZE AND SHORT LIFE SPAN OF MICROCEBUS MAKE IT AN ATTRACTIVE MODEL OF PRIMATE SYSTEMS. A REPRESENTATIVE IMAGE (SEE FIGURE 2) OF A 12-YEAR OLD FORMALIN FIXED MOUSE LEMUR BRAIN SHOWS FEATURES CONSISTENT WITH IRON DEPOSITS LOCALIZED MAINLY IN NUCLEI OF THE HYPOTHALAMUS AND SUBTHALAMUS. THIS DEMONSTRATES THE FEASIBILITY OF IMAGING LIVE ANIMALS UNDER ANESTHESIA.
[pic]
Figure 2. A 12 yr Microcebus brain imaged with MRI (inset) then sectioned and stained for non-heme iron with Prussian Blue stain. There is excellent correspondence between the Prussian Blue stain and MR image intensity. In agreement with MR results, no Prussian Blue stain was noted in the young brains.
In vivo in utero MRI of a day-12 pregnant mouse. Artifacts arising from sample motion are an ever present problem in MR imaging. We investigated this problem by in vivo imaging pregnant mice. A pregnant mouse at day-12 gestation was put into deep anesthesia with an IP injection of phenobarbitol, placed in an Alderman-Grant RF coil, then into our 11.7T MRI instrument. A number of MR datasets were recorded over the next 4 hours. A multislice spin echo protocol taking 10min per experiment was employed. Most of the datasets had significant motion artifacts due, presumably, to diaphragm movement transmitted to the uterus. Some datasets were of high fidelity. One such dataset with no apparent artifacts is shown in Figure 1. In the upper left, a slice through the uterus shows five tightly packed mouse pups. The 15 smaller panels show serial slices through a single pup. Slice position and orientation are shown in the cartoon to the lower right. Ventricles are apparent in panels c-e; convolutions in the snout are seen in f-i; and the digits of the right paw can be seen in panels i and j. Subsequent to this imaging experiment, gestation proceeded normally to term with no obvious ill effects on the mother or pups.
RF & gradient coil optimization. Over the past several years we have explored the characteristics of a number of different types of RF coils with special emphasis on signal/noise performance at 500 MHz, the frequency at which magnetic resonance occurs at 12T. Birdcage resonators (Joseph, 1989; Hayes, 1985; Watkins, 1988), modified saddle coils (Vaslow, 1991), implanted coils (Hollett, 1987), and more novel designs particularly suited to high frequencies (van Vaals, 1990) have been considered. Sadly, at 500MHz coils with diameters larger than ~3cm become difficult to tune due to distributed capacitance effects (Bowtell, 1992). Because RF coil sensitivity is a key parameter in improving the poor NMR signal, it is a key feature of the hardware. For imaging the smaller embryos shown in Figure 2, we used simple solenoid coils and a tune/match network positioned as close as possible to the coil. For samples in the 5-10mm size we used a loop-gap resonator composed of a series of single turn solenoids each tuned to 500MHz. This coil has excellent B1 homogeneity & a high Q. For larger samples, we have made both Alderman-Grant and birdcage resonators. Both have good sensitivity and are straightforward to design and construct. Prototypes of sizes appropriate to imaging both in vivo and in vitro mice brains have been built and function adequately (see Figure 2). Recent work by Dodderll and colleagues has been especially helpful (Eccles et al., 1994; Crozier et al., 1994; Hsu et al., 1995; Mahony et al., 1995; Roffmann et al., 1996).
High resolution macroscopic in vivo imaging. By acquiring multiple individual scans on standard 1.5T clinical scanners, and averaging the MR signal post hoc (see Appendix for details) we have produced an extremely high quality image of a single human subject (Figure 3). This approach will be extended using the high field machine to acquire finer initial resolution and thereby push the limits of in vivo macroscopic MRI in the mouse.
[pic]
Figure 3. Details from N=27 T1 average volume. By combining seven 0.78mm3 and twenty 1.0mm3 volumes, the signal/noise ratio was enhanced beyond that of a single scan, resulting in previously unobserved anatomic detail. Subtleties of nuclear boundaries and fascicles within the thalamus and hippocampus became evident (a) as did brainstem divisions (b). Marked intensity differences within the basal ganglia highlighted their components (c) and interconnections, such as the gray bridges between the caudate and putamen. These gray-gray differences were also clarified within the brainstem (d), and even the small penetrating vessels were resolved (in for instance, the perforated substance (d) or embedded in the insular cortex).
μPET (Molecular Mapping)
Positron Emission Tomography (PET) is a non-invasive imaging technique where positron-emitting isotopes such as carbon-11, nitrogen-13, oxygen-15 or fluorine-18 are attached to biologically relevant molecules and injected in trace amounts into the subject. When the isotope decays, the emitted positron annihilates with an electron in tissue to produce two 511 keV gamma rays which are emitted 180° apart. A PET scanner detects these pairs of gamma rays externally and reconstructs tomographic cross-sections. The intensity of the image at any point reflects the concentration of the isotope (and therefore the tracer of interest). The main advantages of PET are that it is an extremely sensitive technique (picomolar concentrations of tracer can be detected) and that the available isotopes allow just about any molecule of biological interest to be labeled with a positron emitter with little or no modification to the molecule's biochemical activity.
PET is essentially an in vivo analog of autoradiography - indeed it is often referred to as such. The potential advantage of PET however is clear. PET can study the same animal repeatedly, giving investigators the ability to acquire longitudinal studies. While PET may never approach the resolution of autoradiography, it does have sufficient resolution to address certain applications and it has the critical advantage that the animal remains alive to be studied at another time.
μPET is a very high resolution positron emission tomography (PET) scanner developed at UCLA over the past four years and designed specifically for animal imaging (Cherry et al, 199x). The motivation for μPET was to permit repeat, non-invasive studies of biological parameters in vivo. μPET is able to resolve volumes as small as 6 µL (1.8 mm spatial resolution in all directions) and has an absolute sensitivity of up to 200 cps/µCi. Images from μPET are fully quantitative. μPET has been used in a range of animal models, including mice, rats, rabbits and small non-human primates. Over 1000 studies have now been successfully completed on this prototype system using a variety of tracers such as [F-18]fluorodeoxyglucose (FDG), [F-18]fluoroethylspiperone (FESP) and [C-11]WIN35,428 (WIN). Limitations of the current system include the small axial field of view and the limited sensitivity, both of which can be improved by adding more detectors to the system. μPET is a fully 3-D scanner acquiring volumetric data from this imaging field of which can be reconstructed into any arbitrary voxel dimensions and matrix size. Typically, data are reconstructed into 128x128x25 volumes (usually displayed as 25 transverse slices covering the 1.8 cm axial field of view) with cubic voxels measuring 0.7 mm. Each time point produces a raw dataset size of 1.54 Mbytes
Compounds currently used with μPET:
FDG (glucose utilization)
13N-ammonia (myocardial perfusion)
11C-raclopride (dopaminergic neurons - presynaptic)
11C-WIN 35,428 (dopamine transporter)
18F-fluorodopa (dopamine synthesis capacity)
64Cu- labeled antibody fragments (tumor targeting)
18F-fluoroethylspiperone (D2 receptor – postsynaptic; imaging of reporter gene expression) (REF!!!!)
18F-fluoroganciclovir; 18F-fluoropenciclovir (imaging of reporter gene expression) (REF!!!!)
Cryosectioning.
Our preliminary studies have concentrated on development and validation of the technology to collect high spatial and densitometric imagery of whole brain (Toga et al., 1994). During these studies we have sectioned and collected data from human, monkey, rat and mouse brains, sometimes with the entire cranium intact. To accomplish this, we have systematically developed a battery of histological protocols for optimal preparation of frozen specimens. We made a series of engineering modifications to the cryomacrotome to permit accurate image capture directly from the blockface. An automated specimen placement feature, camera, and illumination system were engineered and optimized for consistent and reliable collection of in-register serial image data (Figure 4).
[pic]
CryoImage acquisition. We have built and tested a sectioning and imaging system able to section through the entire neuroaxis of brain and retain accurate spatial dimensions. We have performed preliminary validations using a dial caliper to determine y axis and z axis drift of the cryomicrotome and reconstructed and resampled data volumes to visually inspect any variations in alignment or magnification. These tests resulted in a measured y axis and z axis variation of 0.04% and 0.08%, respectively. Overall illumination was measured using a plain grayfield background during image capture and found to be symmetric, although not completely uniform. Our current illumination system is comprised of fiber optics driven by a Cuda voltage regulated power supply and color temperature matched to the filters of the current camera system. The variations that do remain are corrected using a 3D histogram equalization process developed in the laboratory.
Specimen preparation. Fixed tissue, necessary for some of the histological treatments, may be preferred over fresh frozen material as the intact, fresh frozen brain tends to lose its integrity and shred during sectioning, which we believe is an interaction of the friable or delicate unfixed sections and the intrinsic or a priori matrix destruction caused by freezing artifact in fresh, noncryoprotected large specimens (Duvernoy, 1988). Infiltration of the fixed specimen with appropriate cryoprotectants such as glycerol greatly reduces or eliminates freezing artifact and markedly improves sectioning characteristics (see Rosene and Rhodes, 1990). We have examined the changes in tissue with different cryoprotectant and fixative processing (refs). These can influence the spatial integrity of the reconstructed data.
Acquisition of histological data. We investigated a number of protocols for section retrieval, including adhesive tape (Holmbom et al., 1991; van Leeuwen et al., 1990; Jossan et al., 1991), high molecular weight polymers including polyvinylpyrrolidone (PVP) and polyvinyl alcohol (PVA) (Aaron and Carter, 1987; Hill and Elde, 1990; Fink, 1992), and gelatin (Heinsen and Heinsen 1991; Hine and Rodriguez, 1992). We have used gelatin on the blockface to obtain sections of cryosectioned brain over a broad range of thickness from 50μ up to 900μ (Quinn et al., 1993). We have made appropriate modifications to an engineered roll plate provide reliable collection of quality tissue sections.
[pic]
[pic]
Figure 5. Digitized stained cryosection image. This figure demonstrates a portion of a flatbed-scanner digitization of a 50-micron section (left), collected after cryosectioning (right, showing sagittal reconstruction of coronal sections) and stained for cell bodies.
Spatial Normalization. Two aspects of spatial normalization require particular attention. First, serially sectioned tissue must be realigned to reconstruct the volume from which it was sampled. Second, reconstructed volumes of histologic, cryosection, or histochemical data must be repositioned and/or warped to make them coincident with either an atlas coordinate system or other volumetric data. The ability to compare histologic and molecular maps with in vivo metabolic, gene expression or growth rate data is critically dependent on the accurate alignment of serial tissue sections with a digitally reconstructed specimen (Mega et al., 1997, 1998).
At UCLA, over 15 years of research have been directed at developing tools to transform, deform, and correlate multimodality datasets. These algorithms have been used to integrate 2D, 3D and 4D brain data in a variety of species, acquired from multiple subjects and imaging devices (Toga and Arnicar, 1985; Banerjee and Toga, 1993; Cannestra et al., 1997; Thompson and Toga, 1996, 1997a,b; 1998a,b, 1999a,b,c,d; Mega et al., 1997, 1998; Toga, 1998; Woods et al., 1998a,b, 1999). We have also validated a range of tools to map 2D and 3D histologic, neurochemical and quantitative protein expression data into precise structural register with in vivo MRI and metabolic PET data acquired from the same subject in vivo (Mega et al., 1997, 1998; Cannestra et al., 1998). Additional studies have correlated high-resolution MRI, cryosection and myeloarchitectonic data (Holmes et al., 1996, 1997). In on-going projects, we have synthesized multi-subject brain atlases from a variety of modalities (Toga and Thompson, 1998, 1999; Mega et al., 1999). These include atlases that are dynamic (Thompson et al., 1998, 1999), probabilistic (Mazziotta et al., 1995; Toga and Thompson, 1998, 1999), or specific to a particular diseased subpopulation (Thompson et al., 1998, 1999; Mega et al., 1999; Zoumalan et al., 1999).
Alignment of Histologic Sectioned Material. For histologic data, serial section alignment is the first step in single subject reconstruction. Physical sectioning procedures, required for histological staining, require a superpositioning scheme to re-register the series, prior to comparing it with data from other modalities or with the atlas. Section alignment may also require local deformation or warping transformations (Toga, 1998), to correct for complex patterns of tissue deformation during processing. Two situations fall within the scope of this proposal: the first occurs when a full cryosection volume is available as a reference for histologic alignment. This is the case during atlas construction, as histologic and genetic maps are assembled from subjects that populate the atlas. The second occurs when tissue sections are acquired without a corresponding blockface image. This situation requires multiaxial atlas navigation and computational search tools to identify the optimal alignment of the sections into the atlas.
2D-to-2D Blockface Alignment. Use of blockface imagery greatly eases the problem of registering stained tissue (Mega et al., 1997; Thompson and Toga, 1999). Data collected on the cryomacrotome blockface is in precise spatial register, due to the tomographic nature of the image acquisition (see above). Blockface imagery provides a reference against which other data (from the same specimen) can be compared. Additionally, harvested tissue sections may be cut into small tissue quadrants for regional radioimmunoassay, Western blotting, or ELISA. In recent studies we have linked quantitative results of these assays can then be linked with their original blockface locations for correlation with co-registered structural or metabolic data obtained in vivo (Mega et al., 1997).
In one approach (Mega et al., 1997, 1998, 1999; Thompson and Toga, 1998; Shah et al., 1999; Sanchez et al., 1999), we elastically warped stained sections back into their exact blockface configuration with an algorithm based on the principles of continuum mechanics (Thompson and Toga, 1998). Homologous landmark curves and points were first identified in the distorted tissue and blockface images. A complex 2D deformation vector field is applied to reconfigure the tissue section into the exact shape of the blockface.
Figure 6: Recovery of Change in Brain Tissue due to Post Mortem Effects and Histologic Processing. Warping algorithms based on continuum-mechanical models can recover and compensate for patterns of tissue change which occur in post mortem histologic experiments. A brain section (left), gridded to produce tissue elements for biochemical assays, is reconfigured (middle) into its original position in the cryosection blockface (Mega et al., 1997; Thompson and Toga, 1996, 1998). Note the complexity of the required deformation vector field in a small tissue region (magnified vector map, right) (Thompson and Toga, 1996; Schormann et al., 1996). These data can also be projected, using additional warping algorithms, onto in vivo MRI and co-registered PET data from the same subject for digital correlation and analysis (Mega et al., 1997).
Mathematical Considerations. In (Bookstein, 1989; Joshi et al., 1995; Thompson and Toga, 1996, 1997, 1998; Davatzikos, 1996; Thompson et al.,1998), a general approach is developed for landmark-driven warping of 2D and 3D images, deforming them into register with target images. Landmark points present in one image are forced to match up with their counterparts in the other image. The image to be deformed is considered to be embedded in an elastic block, which is subjected to internal forces that deform it into the exact shape of a target image. In this mapping, internal neuroanatomic boundaries are warped into the configuration of their counterparts in the target brain. The algorithm finds the 2D displacement vector field u(x):R2 -> R2 , mapping one anatomy onto the other, such that the following system of partial differential equations is satisfied:
L(u(x)) + F(x-u(x)) = 0, x in R,
u(x) = u0(x) , x in M0 or M1,
In these equations, L is a differential operator controlling the way in which one anatomy is deformed into the other, and its properties can be used to make the deformation reflect the mechanical properties of deformable elastic or fluid media (Fig. **). Common choices of the differential operator L are the Laplacian Ñ2, biharmonic Ñ4 (Bookstein, 1989; Meyer et al., 1997; Kim et al., 1997) and Cauchy-Navier operator (l+m)Ñ(Ñ·) + mÑ2 (Miller et al., 1993; Christensen et al., 1996; Davis et al., 1997; Thompson and Toga, 1998). The term F(x-u(x)) corresponds to a continuum-mechanical body force distributed though the deforming medium (see Davatzikos, 1996; Schormann et al., 1996; Thompson and Toga, 1998, 1999 for details), and M0, M1 represent sets of anatomical points and curves where vectors u(x) = u0(x) matching tissue regions with their counterparts in the blockface image are known. Prior to making more local adjustments, the algorithm calculates an initial global alignment of the section from the landmark data. Recent studies have favored the use of Procrustes registration over principal axis techniques for automated global alignment (Schormann et al., 1997). When some landmarks are detected automatically, outliers which may adversely affect the global alignment of the tissue section may be eliminated (Black and Rangarajan, 1996). This can be achieved by using iterative closest point or robust point matching (Besl and McKay, 1992; Rangarajan et al., 1996, 1998; Feldmar and Ayache, 1996; Gold et al., 1997; Subsol, 1998), or by using a Rayleigh-Bessel distribution (Schormann et al., 1996), or a Hotelling's T2 statistic (Bookstein, 1989; Thompson et al., 1997; Zhou et al., 1999) on the deformation vectors to reject incorrect landmark pairings.
We have performed extensive cross-validations of automated approaches for section alignment, which include cross-correlative and least-square intensity difference techniques in the spatial and frequency domains (Banerjee and Toga, 1993). In series of coronally sectioned autoradiograms, fiducials inserted orthogonal to the cutting plane (Toga and Arnicar, 1985; Goldszal et al., 1994) were used to evaluate the automated methods for
global alignment.
Local Tissue Transformations. To correct for more local tissue deformations, excellent cross-modality alignment can be achieved by maximizing a measure of the mutual information between the deformed tissue and the blockface as the deformation field parameters are adjusted (Meyer et al., 1997; Kim et al., 1997; Viola and Wells, 1995). For each modality, the performance of intensity-based functionals for automated section alignment will be cross-validated against manual methods using large sets of anatomical landmark curves and surfaces, using methods developed in Thompson et al. (1999). We have recently used such an approach to systematically investigate the performance of the widely-used Automated Image Registration package on data from many subjects and sources (Woods et al., 1993, 1998a,b, 1999; Thompson et al., 1999a,b; Thompson and Toga, 1999).
As the complexity of the alignment transformation increases, automated approaches for deformation recovery may become unstable as the optimization functional becomes non-convex on the deformation field parameter space (Woods et al., 1998). To avoid this, we developed an approach in which anatomical models are created in each dataset. Anatomical knowledge on how landmark features match up is then leveraged to constrain the mapping of one anatomy onto the other. With proper (Dirichlet or infinite) boundary conditions, the matching vector field equation can be solved numerically by finite difference (Christensen et al., 1993; Davatzikos, 1996; Thompson and Toga, 1998), finite element (Gee et al., 1998) or spectral (Christensen et al., 1996) methods.
Although this requires the solution of a system of millions of coupled elliptic partial differential equations, we implemented a fast multigrid algorithm, combined with Chebyshev over-relaxation, which greatly accelerates the computation (Davatzikos, 1996; Schormann and Zilles, 1996; Thompson and Toga, 1998, 1999). In our approach, landmark curve tagging is carried out in a graphical Java Interface that runs on any platform (e.g., UNIX or PC). The software requires only a few seconds of interaction and computation per tissue section (Thompson et al., 1998). The continuum-mechanical model is based on the Cauchy-Navier operator of linear elasticity, and models how real physical tissues deform. Similar tissue alignment procedures have been developed by other groups (Schormann et al., 1995; Davatzikos, 1996; Grenander and Miller, 1998), based on thin-plate splines (Bookstein, 1989; Kim et al., 1997; Meyer et al., 1997), Laplacian or volumetric splines (Davis et al., 1997), Fourier space regularization (Gabrani and Tretiak, 1999), and fluid models based on space-time velocity regularization (Grenander and Miller, 1998; Joshi et al., 1999). For a comparative review of these approaches, discussing the benefits and limitations of each, see Thompson and Toga (1998, 1999).
3D-to-3D Alignment of Cryosection, Metabolic and In Vivo Imaging Data. Each complete cryosection volume, with its realigned tissue and gene expression data, will be subsequently realigned with a high-resolution micro-MRI volume. This volume will be acquired from the same subject in vivo at the final time-point in the MR imaging series. The reconfigured blockface and histologic data therefore represent the specimen's anatomy in its in vivo anatomical configuration, at a resolution obtainable only post mortem. To align cryosection volumes with in vivo anatomic data, a 3D elastic transformation is required, to compensate for post mortem anatomic distortion. First, model surfaces are automatically extracted from each dataset. These model surfaces include cortical, ventricular, hippocampal, callosal surfaces and the internal surfaces of the basal ganglia, many of which can be extracted automatically, in parameterized format, from both MRI and cryosection data (Thompson and Toga, 1996; Holmes et al., 1996, 1997; Zhou et al., 1998, 1999a,b; Pitiot et al., 1999). Internal anatomical correspondences are enforced in the surface-to-surface mappings using a covariant regularization approach specialized for handling cortical and deep subnuclear data (Thompson et al., 1998, 1999a,b,c). Surface-to-surface mappings are extended to the full volume using a 3D continuum-mechanical transformation that matches functionally important interfaces exactly (Thompson and Toga, 1996). Again, where possible, automated image registration approaches (Woods et al., 1998b, 1999) are used first, to recover polynomial deformations between gray-scale cryosection and MRI data. This accelerates the automated extraction of structures that drive the residual local elastic deformation (Thompson et al., 1999; Pitiot et al., 1999; Zhou et al., 1999). In vivo metabolic (PET) data are then aligned to the mutually registered MR, cryo and histologic volumes (Mega et al., 1997, 1998). Whole-brain PET volumes are aligned to the volumetric MR data by optimizing the coefficients of a 6 parameter rigid transformation, based on the ratio image uniformity of the MRI and aligned PET signal fields (Woods et al., 1993). Recent validations have shown highly accurate cross-modality alignment, so long as the optimal algorithm is used for each submodality transformation (Mega et al., 1997).
[pic]
Figure 6. Common space facilitates multi-scale viewing. This figure demonstrates the use of a common coordinate space to permit the viewing of superimposed, multi-scale data from different subjects and modalities. In this case, a low resolution (1.0mm3) in vivo MRI (grayscale) image is overlaid by a much higher resolution (100 micron) cryosection image (hot metal colorscale, see Appendix) from a different post mortem study.
Reconstructions and Visualizations. Anatomic segmentations, descriptions, reconstructions and visualizations have been part of our preliminary efforts also. We can generate 3D models from the data using two different techniques. In the first technique, the anatomic boundaries are defined by the user with contours. Although this is relatively tedious, it does allow the investigator to interpret the image and differentiate the sometimes subtle textures that help separate adjacent structures. The image below illustrates the results of these efforts.
ADD SWANSON and/or 3D model here
The second technique for visualizing reconstructions involves the use of a volume rendering technique. This approach does not explicitly define the surface geometry but can produce 3D models of structure. Its success depends on the degree of contrast attainable with the white matter stains. The resulting models generated using either technique can be resampled using cutaways to expose different planes of section.
Combination too……………..
[pic]
Figure 7. Volume rendering of downsampled cryosection data. To demonstrate the usefulness of even low-resolution cryotome data for anatomic visualization. This figure shows the anterior commissure as it crosses the midline. The figure was prepared by downsampling a 10243 cryosection data set to 2563 then rendering it (interactively) using the hardware texture mapping and an intensity based opacity scale. The anterior commissure can be followed for its entire extent, to the point where it becomes distributed under the globus pallidus. This technique will be extended to allow the interactive realtime viewing of superpositions of complementary information from multiple modalities.
Atlasing
We have produced atlases of several species both in print form containing serial sections and descriptive templates as well as truly 3D digital atlases that have interactive capabilities. The essential element to most atlases is the spatial correspondence between the description of the anatomy (or functional anatomy) and a numeric index. This enables coordinate ↔ nomenclature ↔ anatomy to be equated.
Figure 8. Atlas of the rat. This figure illustrates the product of this digital rodent atlas and the steps required to construct it (right)
Rat. An electronic atlas of the rat was produced using data from the Paxinos and Watson templates reconstructed into 3D (refs). This early prototype included data from multiple animals ‘stitched’ together to form a complete survey of the whole brain. From that set of experiments we recognized the need for sufficient sampling frequency, the need to use either a single animal or a population with appropriate warping for spatial normalization. In a subsequent project we collected a single animal for the creation of a digital web based rat atlas (loni.ucla.edu/data/rat/; Toga et al., 1995). Figure 8 shows a horizontal example from this database.
Monkey. We also have created an atlas of the Nemistrina monkey (Cannestra et al., 199x) that included multiple modalities, specifically MRI, PET, CT and cryosectioned data. These were then spatially normalized relative to one another, placed within a coordinate system, labeled and made available electronically. Animation sequences describing systems of the brain were created as well and have been used for teaching graduate and medical students (Cannestra eta l.,199x). We have also produced atlases in the more traditional print format, including the Rhesus monkey (Paxinos, Huang and Toga, 1999) and the rat (Swanson, 1997) and combination atlases that include both print and electronic components (Swanson, 1999)
Human. Human atlases that describe anatomic detail from multiple modalities have been created too. While these have a different focus from the current project, their construction required that we deal with many of the technical issues necessary for success in creating a useable multimodality, multidimensional atlas of the mouse. ADD HERE
Although this is not an exhaustive list of the essential elements for this project, we have worked on and published in these relevant areas;
1. imaging technology and acquisition
2. imaging gene expression
3. registration, refs
4. reconstruction, refs
5. spatial normalization,
6. coordinate systems,
7. multiple modality correlation,
8. mathematical image analysis and statistics
9. visualization,
10. interaction
11. measurement.
12. Population based image databasing
13. Multispecies atlas dissemination
Mathematics and Computer Science
Reconstruction
Visualization
Intensity and feature (point, curve and surface) based deformation
Imaging
Anatomy
Informatics
Genetics
Genetic (or Molecular or Biochemical) Methodologies
In Situ Hybridization Tissue is frozen, cryosectioned, and gently fixed. Cellular membranes are premeablized, and subsequently probed with a radiolabelled antisense cDNA. This process yields excellent resolution, down to intracellular localization, but it is dificult to quantitate. Northern blot RNA is isolated from a tissue sample, electrophoresed, and immobilized on a nylon membrane. The blot is then probed with a radiolabelled cDNA, exposed on a phosphor storage screen, and quantified on a phosphorimager. This method is considered highly quantitative, but requires a lot of tissue, and therefore has low resolution.
RT-PCR RNA is isolated from a tissue sample, Reverse Transcribed, and amplified by PCR. The cDNA sample is then electrophoresed, immobilized on a nylon membrane, and probed with a radiolabelled cDNA. The hybridized blot is exposed on a phosphor storage screen and quantified on a phosphorimager. This method is also considered fairly quantitative, and requires less tissue than a Northern blot. It is, however, subject to
non-linear amplification due to PCR.
Immunohistochemistry Tissue is frozen, cryosectioned, and gently fixed. Cellular membranes are premeablized, and subsequently probed with an antibody specific to the protien of interest. The protein is visualized with the use of a flourochrome-labeled secondary antibody and flourescence microscopy. This process yields excellent resolution, down to intracellular localization, but it is dificult to quantitate.
Western blot Protein is isolated from a tissue sample, electrophoresed, and immobilized on a nitrocellulose membrane. The blot is then probed with an antibody specific for protein of interest and visualized . The blot is exposed on film, the exposure scanned into a computer by a laser densitomiter, and relative signal strengths compared. This method is considered quantitative, requires little tissue, but must have an antibody to the protein.
Relationship with other Grants
The present application complements other funded projects from our group. The experience gained from the research activities of these grants provides considerable advantage to this project. In particular, there are five funded projects in which the principal investigators participate. These are titled International Consortium for Brain Mapping (ICBM), Digital Representation of Human Brain, Adaptive Algorithms and Quantitative Transformations. The funding agencies for these are shown in Figure 9.
[pic]
The present proposal will benefit from these projects by leveraging algorithm development and technique development with other species and data. Algorithms: Computer science efforts that can be built upon can be found in funded projects from the National Science Foundation and from the NCRR. The NSF grant is focused on the development of warping strategies to compare one brain with another. While the morphology of these data sets is static, subroutines have been developed which can be utilized in the proposed work. The computer science expertise and mathematical background necessary for the development of these warping strategies will be of great assistance to the proposed project. The NCRR grant has set the stage and created the framework for the current application and provides a large battery of algorithm development and software development activities to help ensure the success of this mouse atlas program.
Research Design and Methodology
Design
The overall design of this proposal has the following components;
[pic]
Acquisition, five modalities will be collected including MR, PET, Cryo-blockface images, histology and the anatomic templates
Subject populations, adult mice of the C57BL/6J strain at 3 different weight categories will be collected in all the above listed modalities. Animals at 9 different maturational stages also will be collected corresponding to the Theiler stages xyz. The number of subjects for each of the 12 groups (3 weights and 9 developmental stages) will be determined based upon observed within group variability (PAUL FIX THIS; WHAT IS NEEDED POWER??). In addition, we will collect example subject 1-3 subjects each for the strain XYZ in all modalities.
Volume reconstruction. Volumes for each of the above will be processed (including registration, image processing and JHKJH) and reconstructed.
Spatial normalization. Spatial normalization to equate all subjects and all modalities into a common stereotaxic coordinate system. Procedures for placement of individuals into this system will be defined and tested. Feature based (bony and soft tissue landmarks) and intensity based automatic approaches will be employed.
Anatomic delineations. Anatomic delineations of adult (weight category xx) will create templates identifying individual structures within the coordinate system.
Template transformation. Templates from histology will be transformed to the other modalities and correspondance determined after various types of deformation correction.
Deformation strategies. Deformation strategies to equate different subjects will be developed and tested for their efficacy transforming different weight categories, developmental stages and strains.
Software. Interactive software for the mapping of new datasets, examination of the multimodality, multidimensional atlas will be developed.
Atlas Database. An indexable archive containing the multiple modalities and subjects will be created. Queries of location by coordinate or anatomic name will reference a comprehensive volume and pointing to a place within the volume will produce the coordinate or anatomic name. In addition, the display will be selectable as a single or multiple modalities.
Genetic Expression Maps (GEMS). GEMS will be collected to test the utility of the atlas and associated software.
Acceptance testing. Beta and community testing will be employed to improve the atlas and software throughout and via outreach meetings with the gene mapping and brain mapping communitities.
Development
Multiple Subjects
Additional strain
Multiple modalities (single subjects; in vivo – in vitro)
[pic]
create figure that shows input to the atlas; MR, Cryo, Histo, template, PET
Acquisition
MRI. The overall goal of component of the project is to implement MR imaging as a facile means of obtaining 3D anatomical information in mice. Toward this end, much of the necessary MRI hardware is already in place. We plan modest MRI probe design changes optimized for imaging the mouse and a significant MRI console upgrade. We will investigate and implement imaging protocols that provide the maximal relevant information in the minimal time. During the all phases of this project we will ensure that the instrumentation is appropriate for all phases of this project. In this Section we discuss:
• probe design approach,
• imaging protocol development for in vivo mouse MRI at high fields,
• mouse model systems we will use as initial test cases.
Probe Design for Magnetic Resonance Imaging Microscopy. The MRI probe contains the RF coil with ancillary circuitry, magnetic field gradient coils, life support and monitoring, and plumbing to the outside world. To acquire an image the specimen is placed inside the probe and the assembly is placed inside the large static magnetic field. Thus, optimization of all probe characteristics is essential to secure good images and stable preparations. Design considerations involve ensuring the viability of the in vivo samples being imaged, as well as, achieving the necessary signal-to-noise ratio (SNR) and resolution. Due care will be taken to minimize the potential for vibration when integrating all the pieces of the probe. The trial-and-error method works well here because these RF circuits are conceptually straightforward, relatively simple to build, and easy to test. We will concentrate on adjusting the following sections:
• Specimen environment and monitoring,
• RF coils,
• Gradient coils.
Specimen Environment and Monitoring. The primary concerns here is to provide temperature control, maintain humidity, and control oxygen and gaseous anesthesia (typically vaporized isoflurane) levels for the in vivo specimens. We have successfully accomplished this in our 89mm vertical bore system and will use this instrumentation in the present project. An all plastic animal constraint and positioning setup has been specially built for our small bore horizontal magnet. It allows for the precise repositioning of the specimen in the instrument. This is essential for repeated imaging of the same specimen (development) and for similar positioning of different specimens.
Radio Frequency Coils. It is through the RF coil that we excite the proton spins in the sample and detect the ensuing MR signal. Good sensitivity, a homogeneous B1 field, and efficient coupling with the sample translate directly into high fidelity images. We have a number of solenoidal, loop-gap, surface, Alderman-Grant type, and birdcage coils of different sizes on hand. When necessary (e.g. to maximize filling factor) we will construct additional coils. Typically, we calculate coil characteristics (e.g. B1 field uniformity) using MATHEMATICA, make a mask using a standard CAD package, etch the coil pattern on a copper-teflon laminate, then construct the coil with the etched copper-teflon laminate and appropriate nonmagnetic capacitors. We have found this a convenient and relatively inexpensive way of making prototype coils. An HP 8752A Network Analyzer will be used to test the performance of the isolated coils, coils plus matching/tuning circuits, and the whole probe. This will allow is to conveniently and efficiently evaluate the individual elements as well as the combined assembly.
Gradient Coils. ???????????????????
Acustar gradient set – ID, max strength & rise time, max sample size
Micro2.5 gradient set - ID, max strength & rise time, max sample size
No additional gradients planed as these fit mice just fine.
Imaging Protocols. The are basically three types of software in MR imaging: i) pulse sequences, ii) image reconstruction programs, and iii) image presentation & analysis routines. When implementing imaging pulse sequences at high fields, one must keep in mind that many phenomena (e.g. magnetic susceptibility effects, chemical shift dispersion, relaxation times) are quite different at 11.7T than at typical clinical field strengths of 0.5-1.5T (Sharp, 1993; Bowtell, 1992; Bowtell, 1990). To gain as much structural and anatomical information as possible from our specimens, we will employ several different pulse sequences aimed at obtaining images where contrast arises from different physical aspects of the sample (e.g. proton density, relaxation times, magnetic susceptibility, diffusion). For our in vivo specimens the images must be acquired in a relatively short period of time. Short recycle time spin echo (SE) sequences with multiple echo acquisition will allow us to obtain both T1 and T2 weighted images simultaneously. Diffusion weighted and in selected cases diffusion tensor imaging will be used to gain information about myelination. We will implement these algorithms and will assess the image characteristics versus time needed to acquire the data in order to arrive at the most appropriate sequence(s) for collecting the information needed in this project.
Pet.
The goal is to use a reporter gene based method which permits the location, magnitude and persistence of gene expression to be measured quantitatively in the intact mouse.
The basis of the method is to use a reporter gene whose product, rather than being fluorescent, can trap a positron-labeled probe (PET reporter probe). Possible systems under investigation include a receptor based system, where the gene for the dopamine receptor is used as the reporter gene, and an enzyme based system where the herpes simplex virus type 1 thymidine kinase (HSV1-tk) gene is the reporter gene. In the receptor mediated system, [18F]fluoroethylspiperone is used as a PET reporter probe, and is specifically bound only to cells expressing the dopamine receptor. Outside of the striatum, only cells expressing the reporter gene will bind this compound, and therefore its uptake as measured by PET can be related to expression of the reporter gene (MacLaren et al, 1999). The enzyme mediated system uses [18F]-fluoroganciclovir (FCGV) as the PET reporter probe. In the presence of HSV1-TK enzyme, FCGV is phosphorylated and trapped in cells. Thus, the uptake of FGCV can also be related to expression of the reporter gene (Gambhir et al, 1999). To facilitate quantitative and meaningful studies in the mouse, a new high resolution PET system, microPET, has been developed which provides isotropic 1.8 mm resolution images (Cherry et al, 1997).
Initial validation experiments have been carried out by adenoviral delivery of the reporter gene into mice via tail vein injection. In both PET reporter gene/reporter probe systems, it has been possible to demonstrate a quantitative correlation between message levels of the reporter gene and uptake of the PET reporter probe in the microPET scans (Gambhir et al, 1999; Maclaren et al, 1999). An example of a study with the HSV1-tk PET reporter gene, correlated with digital whole body autoradiography, is shown in figure X.
The plan now is to measure the espression of genes of interest using a single promotor to drive both the PET reporter gene and the gene of interest as a bicistronic message, with an internal ribosomal entry site to facilitate translation of both proteins from a common message.
FIG HERE
Figure X caption: MicroPET and digital whole body autoradiography (DWBA) images of mice after adenoviral mediated delivery of the PET reporter gene. Swiss-Webster mice were injected via the tail vein with 1.5 x 109 pfu of control virus (a) or 1.5x109 pfu of AdCMV-HSV1-tk virus (b). For each mouse, a whole-body mean coronal projection image is displayed on the left. The liver outline in white was determined from both the FCGV signal and cryostat slices. The second images from the left are coronal sections, approximately 2 mm thick, from the microPET. After their PET scans, the mice were killed, frozen and sectioned. The next images are photographs of the tissue sections (45 µm thickness) corresponding to approximately the midthickness of the microPET coronal section. The images on the right are DWBA of these tissue sections. The color scale represents the FGCV %ID/g. Images are displayed on the same quantitative color scale to allow signal intensity comparisons among them.
Cryosectioning of frozen specimens. Three dimensional volume data sets of mouse brains will be reconstructed from high–resolution digital serial images of cryosectioned specimens. Whole head and brain specimens will be histologically prepared for sectioning by en bloc fixation and cryoprotection against freezing artifact.
Specimens are sectioned on a large industrial cryomacrotome (PMV Stockholm, Sweden) using a hardened steel knife. We have modified the cryomacrotome to enable quantitative electronic image capture. It is now equipped with a digitally controlled camera and color-balanced fiber optic lighting which is integral with the hydraulic descending knife apparatus. The motorized sledge of the cryotome has been modified to include an automated stop feature placing the specimen directly under the camera in a consistent manner prior to each image capture. These modifications to the PMV cryomacrotome provide blockface imaging technology, allowing capture of in–register serial images at constant magnification throughout the whole–head sectioning process. Figure 12 describes this device.
Whole head and brain specimens are automatically cryosectioned (-20˚C) in 50–100 micro increments. Images of the blockface are captured every 100 microns for most applications. This provides the appropriate relative match of pixel dimension in the vertical and the planar axes. Digital images are transferred to the SGI supercomputer system for multidimensional image processing and archival storage.
Image Processing. Digitization of the block face is accomplished using a software program designed by our group. High resolution color images are captured directly from the blockface and projected onto a color display monitor before being saved digitally. Magnification, illumination, and color balance are held constant throughout the capture and sectioning process. Registration of serial images (x,y values) is maintained by a mechanical feature which automatically positions the specimen underneath the camera. Distance between serial images (z value) is determined by the interval in microns between captured blockface planes.
Histology. To construct an accurate 3-D computer graphics model of the brain from histological sections, it will be necessary to (a) obtain serial sections from a single brain, and (b) obtain these sections from one block-that is, by cutting a whole brain from the rostral tips of the olfactory bulbs to upper segments of the cervical spinal cord. Tissue sections will be taken every x microns. The specific corresponding blockface image will be noted for subsequent 2D deformation correction. As noted above, histological processing can induce significant distortion relative to the relatively representative geometry of the frozen blockface.
Anatomic Description. Describing individual structures will be accomplished using one or more of several strategies developed in our laboratory and adapted for this project. We will do this by manual segmentation, semi-automatically using templates published or developed by us using manual segmentation of other data, density gradients from stained or MR contrast, or using purely probabilistic approaches that are based on location (Mazziotta et al., 1995).
Anatomy-delineation: Neural structures (including cell groups, fiber tracts, and gross anatomical features such as the ventricles) are determined under the microscope using bright- and darkfield illumination of serial, Nissl-stained sections. Different features in the brain have been identified with varying levels of accuracy in the primary literature, and the best delineations rely on the widest array of neuroanatomical information, including cytoarchitecture, chemoarchitecture, and connections (neural outputs of particular cell types within an area, and neural inputs)-information that can be related to the features observed in Nissl-stained sections (Swanson, 1998--second edition of the atlas). Because very little experimental neuroanatomical research has been carried out in the mouse, the vast majority of its structural parcellation must rely on information obtained in the rat. Exactly how similar the architecture of rat and mouse brains are remains to be determined.
Anatomy-nomenclature: Neuroanatomical nomenclature has evolved over more than two thousand years, and for historical reasons it is not logical and internally consistent-like any language. And because there is still a great deal to learn about the structural organization of the brain, neuroanatomical nomenclature must remain flexible. Having said this, there are nevertheless sophisticated and complex conventions that are associated with the refinement of biological nomenclature and taxonomy that involve historical precidence, internal consistency, and so on (Swanson, 1998). We have developed a neuroanatomical nomenclature for the rat that is based almost entirely on references to the primary neuroanatomical literature, is internally consistent for both the adult and the developing central nervous system, and is applicable in so far as possible to mammals in general (Alvarez-Bolado and Swanson, 1996 [development atlas]; Swanson, 1998).
Atlas Templates. In preliminary work for this project, we worked on the segmentation of neuroanatomic structures within 3D data sets using anatomic templates from published atlases like that of Talairach (Talairach and Tourneaux, 1988), Paxinos (Paxinos and Watson, 19xx) and Swanson (Swanson, 19xx). These efforts define a nomenclature to coordinate relationship. Once the model has been placed into the stereotactic coordinate system, as described above, the data set will be resampled so that a given plate from the reconstructed model is identical to a plate described in the atlas. The anatomic template will then be overlaid upon the histological image, such that anatomic boundaries can be seen. Fitting the anatomic templates provided by published atlases to data will greatly increase the number of structures that can be labelled. Digital reconstructions of stained tissue (using blockface images for registration and deformation) will be used to provide additional structural information to more completely map the brain.
Anatomic templates will be prepared by tracing photographs of histological sections. These maps are interpretations of the outlines of structures delineated in the brain, and are named according to criteria discussed above. In our case, digitized photos of brain sections (scanned at 300dpi) are placed in Adobe Illustrator 8, and maps or templates are drawn as transparent overlays (layers). These drawings are vector-based, and as such are essentially infinitely scalable, and can be printed at very high resolution. This approach has the great advantage that an infinite number of perfectly aligned transparent overlays can be created, each of which contains some type of neuroanatomical data (e.g., the results of pathway tracing experiments, immunohistochemistry, or in situ hybridization). These data overlays can be displayed in any order and combination, and they can form the basis of a graphical database of neuroanatomical information (Dashti et al., 1997).
This approach to atlas production-computer graphics templates displayed over digital photographs of brain sections-has the great advantage that other interpretations of structure can also be superimposed over the brain sections. That is, alternative interpretations are easy to accommodate. The basic principle is that a coordinate system is constructed for the brain itself-that is, for the photographs of brain sections. Then, different names (i.e., aliases) can be assigned to the same coordinate(s), and these names are indexed by reference to the photograph itself (or, actually, to the set of coordinates defining the border of structures delineated in the maps or templates. It is also important to define how the coordinate system in the histological sections is related to a coordinate system in the in vivo brain (i.e., from MR image). Based on the MR and histological sections, it is possible to construct a 3-D cartesian coordinate system for the brain.
Genetic Expression
Anatomy
Delineation
Nomenclature
Template
Coordinate System/Stereotaxis
Maps and Models
Warping
|Modality |( |Modality | |Registration Warp Type | |Rationale |
|1. Histo |( |Cryo | |Fluid/intensity/feature | |Cryo is in vitro spatial gold STD but has |
| | | | | | |insufficient anatomic detail; Histo is distorted|
| | | | | | |due to processing |
| | | | | | | |
|2. Cryo |( |MR | |Fluid/ intensity/ feature | |MR is in vivo spatial gold STD |
| | | | | | | |
|3. Template |( |Histo | |Histo | |Histo is anatomic gold STD |
| | | | | | | |
|(3-4.) PET |( |MR/Cryo | |Affine | |All are tomographic so no local deform; PET has |
| | | | | | |lowest resolution |
Warping Data Across Modalities, Across Subjects and Across Time
Elastic image registration, or warping algorithms, calculate a deformation field that reconfigures one brain to match another. Depending on whether the datasets being matched are (1) from different imaging modalities, (2) from different subjects or strains, or (3) from different time-points, warping algorithms have a powerful range of applications.
The 3 major objectives are:
1. Data Fusion across Modalities and Across Subjects (see Background and Significance, page **). Histologic, gene expression and biochemical maps are reconfigured in 2D into their blockface configuration, correcting tissue distortions due to staining. A subsequent 3D-to-3D warp reconfigures cryosection and associated histochemical data into register with micro-MR and co-registered genetic and metabolic PET data from the same subject in vivo. As appropriate (Thompson and Toga, 1999), landmark-driven, model-driven or intensity-based warping will be applied to histologic, cryosection, MR and PET data from additional subjects to register them with an atlas coordinate system or other volumetric data.
2. Population-Based Atlasing: Mapping Within and Between Strain Variability. Warping algorithms create detailed maps of anatomic differences between subjects, at a given developmental stage. When applied in a probabilistic framework, warping algorithms can measure anatomic variability within and between strains, generating quantitative maps of the magnitude and principal directions of variation. Algorithms that average geometric and intensity features across subjects will be used to generate well-defined average templates of anatomy at each time-point in the atlas. Atlas templates will be constructed in the form of volumetric MR and cryosection data, along with sets of 3D structural models. Both models and image templates can be generated in their mean anatomical configuration, using group-specific atlasing methods introduced in Thompson et al., 1999a,b (see Appendix; cf. Grenander and Miller, 1998, for a similar approach).
In (Thompson et al., 1999) we introduced several computational methods for population-based averaging of anatomy. These pattern-theoretic and shape-theoretic approaches (Grenander, 1976; Bookstein, 1989; Miller and Grenander, 1994; Thompson and Toga, 1999) treat geometric and intensity variation separately, and encode inter-subject differences in brain structure. A set of high-dimensional elastic mappings are calculated, based on the principles of continuum mechanics, matching the anatomy of a large number of subjects in an anatomic database (Thompson et al., 1996, 1997; cf. Haller et al., 1997; Csernansky et al., 1998; Grenander and Miller, 1998). These mappings generate a local encoding of anatomic variability, and are used to create a crisp anatomical image template with highly-resolved structures in their mean anatomical configuration.
Probabilistic Anatomical Maps. Probabilistic maps of structure will be invoked to adjust for the effects of anatomic variability when sampling functional imaging and genetic attribute data across subjects (Dinov et al., 1999). They will also be leveraged to provide anatomical prior information for automated structure labeling, parameterization and modeling algorithms (Pitiot et al., 1999; Zhou et al., 1999). If warping approaches are used to measure 3D anatomic differences across subjects, anatomic differences between strains and population subgroups can be mapped. At each time-point, major anatomical systems will be modeled in the micro-MR and co-registered cryosection data using parametric surface meshes (Thompson et al., 1996a,b,c, 1997a,b,c, 1998a,b; Holmes et al., 1996; Mega et al., 1997, 1998). This computational approach provides models that can be compared, averaged or combined across subjects. Surface models can also be measured to provide a variety of morphometric statistics (e.g., surface areas, volumes, complexity, surface curvatures, and other descriptors; Thompson et al., 1996, 1998). The resulting models can also be rendered, visualized graphically and animated (Thompson and Toga, 1997), allowing between strain anatomic differences and structural variations to beillustrated directly.
Genotype vs. Phenotype. Methods to compare probabilistic information on brain structure from different subpopulations are under rapid development, and include approaches based on random tensor fields (Thompson and Toga, 1997a,b, 1998; Thirion et al., 1998; Gaser et al., 1998; Cao and Worsley, 1999), singular value decomposition and ManCova (Ashburner et al., 1998), shape-theoretic approaches (Bookstein, 1997), stochastic differential equations (Christensen et al., 1993) and pattern theory (Grenander and Miller, 1998). By encoding patterns of anatomic variation across subjects, we will determine the effects on brain structure of embryonic age, body weight, brain weight, sex, strain and other knockout-specific genetic factors (Thompson et al., 1999). Multivariate analysis of covariance will be used to identify the patterns of linkage between shape variations in anatomy and other endogenous or genetic factors. At a gross anatomical level, this will facilitate the exploration of relationships between genotype and phenotype in C57BL/6J and other strains.
2. Warping across time to quantitate development, measure growth rates, generate interpolated atlases, and compare data at different time-points.
Changing Morphology. Given the dynamically changing anatomy of the developing mouse, powerful computational tools are required to compare data at different time-points, measure growth rates, and generate interpolated atlases. Interpolated atlases will be created to optimally represent brain structure and its associated genetic and functional attributes at time-points in between those when data is acquired. Several years of research have been directed towards the issues involved in representing a dynamically changing morphology in rodent, Macaque, and human pediatric brain data (Toga et al., 1996; Thompson et al., 1998, 1999; Sowell et al., 1999a,b; Blanton et al., 1998). More recently, we developed tools that measure local growth rates during brain development. These tools also enable the projection of brain maps from one time-point to the next.
A Dynamic Atlasing Framework. The dynamic atlas will be developed in two stages. First, high-resolution MR data from one subject acquired at multiple time-points will be used to generate a sequence of non-linear deformation fields. Based on these fields, in vivo growth rates will be quantified (Thompson et al., 1999). In vivo genetic and metabolic PET data will be correlated with growth rates to clarify the relation between gene expression and the underlying dynamics of brain growth. Additionally, metabolic and in vivo genetic data will be projected forward and backward across time to enable multiple time-point correlations.
Once a single subject has been mapped at all time-points, repetition of this procedure in additional subjects will enable us to generate an average anatomical representation for the group at every time-poimt (Thompson et al., 1999). By creating a series of average anatomical templates for the group, the non-linear registration of these templates across time will be used to identify the generic features of growth in the group, and their relation to instantaneous gene expression patterns and metabolic data across subjects and strains. Once registered, maps of growth rates are simply another form of attribute data linked to the underlying anatomy. These signal fields can therefore be averaged or subjected to principal component or principal deformation analyses (Bookstein, 1989; Thompson et al., 1999). Additional multivariate approaches, linking deformation field variance with exogenous variables (cf. Joshi et al., 1999; Thompson et al., 1999; Davatzikos and Resnick, 1998; Bookstein, 1997) will be used to reveal growth patterns that are characteristic of a particular strain or embryonic phase.
Intermediate Brain Atlases. To generate intermediate atlases, two approaches will be evaluated. After appropriate intensity normalization and preprocessing, linear interpolation of atlas geometry and its constituent maps will be used to generate intermediate atlases. To do this we will use the maps acquired at adjacent time-points and the appropriate percentage of the deformation field required to register them. We have used this approach in the past to create animations of rodent development. These animations were based on surface models of cortical anatomy acquired at multiple time-points, and a distance field blending model (Absher et al., 1994; Payne and Toga, 1996; Toga et al., 1996). A second, more adventurous approach recognizes that structure growth is unlikely to be linear between sampled time-points, since this would create discontinuities in the predicted growth rates at the time-points where data is acquired. By using space-time regularization (Grenander and Miller, 1994; Joshi et al., 1997; Dupuis et al., 1998; Thompson and Toga, 1999) a smoothly differentiable growth rate field can be recovered which is consistent with all the anatomical models in the series. This velocity field for structure growth is unique in the sense that it minimizes a measure of the irregularity of growth rates over the space of all growth profiles that are consistent with the imaging data. For predicting interpolated brain maps, the advantages of this approach will be tested over the simpler, linear approach by leaving out (jack-knifing) sample data sets. The accuracy of each model in interpolating growth or other brain maps can therefore be determined, by assessing how each omitted brain map is predicted by each model from the remaining data, using least-squares measures for each imaging device as an approximation metric.
3D Models. While 2D parameterization supports the analysis of surface changes, fully 3D volumetric changes will be captured using model-driven 3D image warping (Thompson and Toga, 1996). Modeling morphometric changes also requires 3-D methods to quantitate and visualize the rate and extent of the complex growth, pruning or atrophic processes taking place throughout the brain. 3D reconstruction techniques for representing the internal and external geometry of the brain will provide the foundation to model dynamic changes in its cellular
architecture. Manipulation of the morphometric models with image warping techniques will provide a mathematical representation of the changing biological system. To guarantee biological as well as computational validity, these warping strategies must be specifically designed to handle neuroanatomic data in 3 and 4 dimensions. This work will be based on our recently developed algorithm for elastic warping of brain image data (Thompson and Toga, 1996). Anatomic constraints are used to calculate a deformation field that reconfigures the anatomy of one subject into the shape of another. This algorithm provides detailed, quantitative 3D maps of the anatomic differences between one subject and the other Automated warping approaches (Woods et al., 1998b) will be developed in parallel, allowing cross-validation of independent methods (Specific Aim 4).
Warping of one scan onto another acquired at a later time-point can be regarded as finding a mathematical transformation which warps the 3D rectilinear lattice of the original scan into the configuration of a curvilinear mesh which threads through the target scan. This curvilinear mesh describes the complex profiles of dilation, contraction and shearing at a local level in the anatomy, and can be regarded as a 3D parameterization of the later-stage anatomy.
4D Models. Concatenation of a series of warps representing structural change over time will result in 4D parameterization of neuroanatomic change for a given subject in its full spatial and temporal complexity. In practice, the 4D parameterization will consist of a variety of 4D maps of neuroanatomic change. Different features of the local change will be represented by animating scalar, vector or tensor maps of individual deformational characteristics, depending on which attributes are most instructive to emphasize.
Tensor Maps. Tensor maps will be used to show multidimensional quantities characterized by a large set of interdependent parameters at each point in space. Examples include diffusion tensor maps, which encode the magnitude and principal directions of intracellular diffusion processes, and covariance tensor maps (Thompson and Toga, 1997) which encode confidence limits and principal directions of anatomic variation at each anatomic point. Deformation processes recovered by our warping algorithms will be described by a variety of tensors, reflecting the magnitude and principal directions of dilation or contraction, the rate of strain, and the local curl, divergence and gradient of flow fields representing the transformation. In contrast with prior approaches based on rigid registration, each of these tensor maps provides a local measure of structural change.
Figure 12: Connected Surface Systems used to Drive the 3D Warp. Deep surfaces include: the anterior and posterior calcarine (CALCa/p), cingulate (CING), parieto-occipital (PAOC) and callosal (CALL) sulci, the Sylvian fissure (SYLV), the superior and inferior surfaces of the rostral (VTSs/i) and inferior horns (VTIs/i) of the right lateral ventricle. Ventricles and deep sulci are represented by connected systems of rectangularly-parameterized surface meshes, while the external surface has a spherical parameterization which satisfies the discretized system of Euler-Lagrange equations used to extract it. Connections are introduced between elementary mesh surfaces at known tissue-type and cytoarchitectural field boundaries, and at complex anatomical junctions (such as the PAOC/CALCa/CALCp junction shown here). Color-coded profiles show the magnitude of the 3D deformation maps warping these surface components onto their counterparts in a scan from an age-matched normal subject. Color reproductions of this figure can be found in the appendix.
Figure 11: Inter-Modality Warping: Mapping 3D Digital Cryosection Volumes onto 3D MRI Volumes. The result of warping a 3D cryosection image (lower left) into the shape of a target MRI anatomy is shown in (lower right), with cortical landmarks of the target anatomy superimposed. Note the reconfiguration of major occipital lobe sulci into the shape of the target anatomy. Registration of critical lobar, sulcal and cytoarchitectural boundaries is only possible with a high-dimensional warping technique (Thompson and Toga, 1996; Christensen et al., 1996).
Due to the tensorial nature of the maps of growth and atrophy, one approach to calculating these maps will be to use continuum-mechanical models for matching 3D images, pioneered for brain image registration by (Christensen et al., 1993; Miller et al., 1993). These continuum models will be adapted so that they can be driven by structure models with a variety of rectilinear and spherical parameterizations (Davatzikos, 1996; Thompson and Toga, 1996, 1997, 1998). Continuum-mechanical registration algorithms inherently invoke tensor descriptions of structure dilation, contraction, divergence and shearing. Fluid registration models will also be investigated, which incorporate the concept of deformation velocity, rate of strain tensors, and
concatenation of deformation fields through the relationship of material differentiation (Christensen et al., 1996). In addition, computation of warping fields will also be carried out using differential operators which rely on small deformation assumptions (Christensen et al., 1996), such as the bi-harmonic operator. Different regularization approaches may offer specific statistical advantages in the analysis of the resulting deformation maps (Bookstein, 1997).
Probability Distribution
DataBase
Braintree
Multiple Subjects
Additional Strain
The additional strain used should be either a commonly used strain, such as BALB/c, or a highly informative one, such as Spret/Ei (shows the greatest sequence variation from C57BL/6).
Genetic Expression
Gene expression will be studied in a number of ways. Measurements of mRNA from tissue will be done by cytochemical methods such as in situ hybridization, with verification by more traditional electrophoretic methods such as Northern blot and RT-PCR. Measurements of protein expression from tissue will be done by cytochemical methods such as immunohistochemistry, with verification by Western blot.
Software Development
A critical product of this research will be the creation of software. Each of the participating institutions has many years of experience in algorithm research, software development and distribution. The software development that will form the basis of this research project will build upon an extensive library of subroutines and programs. We have created programs that have serviced a wide variety of laboratories, nationally and internationally. We have built code to work on many hardware and software platforms and distributed source code and documentation via the Web and on magnetic/optical media.
Platform Migration. Although the present plans hold for us to develop the programs that will create and interact with the atlas on unix workstations. Many of the interfaces will be built using Java applets (see preliminary results) so that other platforms can utilize the product of our development. A trend that is important to the present application is platform migration. The downward pressure of sophisticated hardware/software solutions towards cheaper, smaller and faster platforms has been a rewarding force in the field, leading to increased use by scientists of technologies and software earlier considered proprietary by military, movie industry and other privileged groups.
To facilitate the porting of programs between platforms and to present a unified interface, we have adopted several open standards for all programs. These standards are shown in the appendix:
Computer Programs
Overall design
Interface
Time Line
Recognizing the need for a rapid prototype development, we will develop a single animal complete atlas of the mouse within the first year. It will not be complete and will not have the final level of anatomic detail in the delineations. But it will enable its use and the overall design to be demonstrated. As noted in the Administration section, we plan on several venues for the evaluation of our prototype atlas. These include an advisory board, the presentation of the atlas at NIH sponsored multidisciplinary meetings and via publication in traditional print as well as web based media. These activities will continue throughout the life of this project. Thus the first year of effort will concentrate on the creation of a framework and a single animal representative. Year 02 will focus on methodological and procedural refinement along with more complete delineations and software development. We will begin testing the system with our own gene expression data. Year 03 will see acquisition of subjects at different developmental stages and the addition of a larger n for all groups. We will also begin beta testing of the software for insertion of data into and interaction with the atlas. In year 04 we will begin distribution of the atlas and accompanying software. In year 05, the comparison of additional strains, the completion of the developmental delineations and the calculations of the probabilities will be achieved.
Alpha – beta dissemination
I. Administration
The administrative plan for orchestrating the different aspects of this project will require monthly meetings, either in person or on the telephone, with each of the principal investigators from each of the institutions. In addition, we will meet in person, going over the results and development of technology every two or three months at 1 of the 3 sites. Finally, as described elsewhere in this proposal, we intend to exchange technical personnel to implement the technologies in each of the other sites.
Administering multisite consortium efforts has been a strength of UCLA, as the Human Brain Project grant entitled "International Consortium for Brain Mapping" is dependent, in large part, upon the cooperative spirit developed among the participants, as well as the administration of investigators at geographically distant sites. Given the fact that the 3 institutions involved in the present proposal are in close proximity geographically, we are confident in our ability, as already demonstrated, to work together cooperatively.
(multisite subcontracts, outreach to genetics community, annual meetings)
Advisory board??
E. Human Subjects
There are no human subjects in this project.
F. Vertebrate Animals
We will use the C57/BL6/J mouse. IRB approvals for all protocols will be obtained from the respective institutions, most notably, UCLA and Caltech. Animals are treated using the veterinary care services of UCLA and Caltech, respectively, and approval notices for the UCLA aspects of this project are under review, as noted on the face sheet. Euthanasia will be accomplished by overdose of anaesthetic followed by perfusion fixation. These methods are consistent with the recommendations of the Panel on Euthanasia of the American Veterinary Medical Association.
[pic]
H. Consortium/ Contractual Arrangements
The organizational structure of this project involves participation from 3 sites, UCLA, CalTech and USC. Investigators of each institution, Drs. Toga, Jacobs and Swanson respectively, already have conducted preliminary experiments together, confirming their ability to efficiently work together towards a common cause. For example, animals from UCLA have been imaged at CalTech and algorithms developed at UCLA have been utilized by USC. Anatomical nomenclature prescribed by USC has been used by UCLA and cooperative anatomic studies between these two institutions have been conducted in several species. Thus, the overall cooperative spirit of this project has already been proved and resulted in several abstracts attesting to the successes of these research activities. Secondly, the three sites are in relative geographic proximity. While in the Los Angeles Basin area, distance is usually measured in time, rather than miles, the commute from site to site usually can be accomplished within half an hour. Thus, students, staff as well as faculty, can be exchanged among sites on a day to day basis and have already done so, as described above. In order to extend this cooperative venture, we have made arrangements to exchange students and staff for longer periods of time, so that they may enhance the cooperation, learn new aspects of the project and help implement a particular technical component at another site. It should be noted that the lead investigators of each site have a long standing personal friendship of too many years to acknowledge.
Thus, this is an integrated research project involving 3 different sites and several investigators. The product of this cooperative venture will be greater than the sum of the parts, resulting in a comprehensive atlas of the C57/BL6/J mouse describing its developing and adult nervous system in vivo and post mortem.
Budget
Subcontracts are included in this proposal, where the budgets for each site are clearly described in the budget section, and their justifications found earlier in this proposal. The lead site, UCLA, also includes overhead costs for the distribution of these funds.
G. References
Aaron, J. E., and D. H. Carter (1987) Rapid preparation of fresh-frozen undecalcified bone for histological and histochemical analysis. Histochem Cytochem 35:361-369.
Aber. J. D. And Mary E. Martin, High Spectral Resolution Remote Sensing of Canopy Chemistry, 5th Annual JPL Airborne Earth Science Workshop, JPL 95-1, vol. 1, pp. 1-4, 1995.
Aguayo, J. B., S. J. Blackband, J. P. Wehrle, J. D. Glickson and M. A. Mattingly. (1987). NMR Microscopic Studies of Eyes and Tumors with Histological Correlation. Ann. NY Acad. Sci. 508: 399-413.
Banerjee, P.K. and Toga, A.W. 1994 Image alignment by integrated rotational and translational transformation matrix. Physics Med. Biol. 39:1969-1988.
Barnett EM; Evans GD; Sun N; Perlman S; Cassell MD. Anterograde tracing of trigeminal afferent pathways from the murine tooth pulp to cortex using herpes simplex virus type 1. Journal of Neuroscience, 1995 Apr, 15(4):2972-84.
Basser, P. and C. Pierpaoli, Microstructural and Physiological Features of Tissues Elucidated by Quantitative-Diffusion Tensor MRI, J. Magn. Reson. B 111, 209 (1996).
Basser, P. J. et al., Estimation of the Effective Self-Diffusion Tensor from the NMR Spin-Echo, J. Magn. Reson. B 103, 247 (1994).
Beach TG; McGeer EG. Tract-tracing with horseradish peroxidase in the postmortem human brain. Neuroscience Letters, 1987 Apr 23, 76(1):37-41.
Beaulieu, C, P. S. Allen, Determinants of Anisotropic Water Diffusion in Nerves, Magn. Res. Med. 31, 394 (1994).
Beaulieu, C., P. S. Allen, An In vitro Evaluation of the Effects of Local Magnetic- Susceptibility-Induced Gradients on Anisotropic Water Diffusion in Nerve, Magn. Res. Med. 36, 39 (1996).
Beaulieu, C., P. S. Allen, Water Diffusion in the Giant Axon of the Squid: Implications for Diffusion-Weighted MRI of the Nervous System, Magn. Res. Med. 32, 579 (1994).
Belliveau, J. W., D. N. Kennedy, et al. (1991). Functional Mapping of the Human Visual Cortex by Magnetic Resonance Imaging. Science 254: 716-719.
Blumich, B., and Kuhn, W. (1992). Magnetic Resonance Microscopy. New York: VCH Publishers.
Boardman, J.W. and F. A. Kruse, Automated Spectral Analysis: A Geologic Example Using AVIRIS Data, North Grapevine Mountains, Nevada, Tenth Thematic Conference on Geologic Remote Sensing, Environmental Research Institute of Michigan, Ann Arbor, pp. I-407-418, 1994.
Boardman, J.W., Automated Spectral Unmixing of AVIRIS Data Using Convex Geometry Concepts, 4th JPL Airborne Geoscience Workshop, JPL Publication 93-26, pp. 11-14, 1993.
Boardman, J.W., Inversion of Imaging Spectrometer Data Using Singular Value Decomposition, Proceedings, 12th Canadian Symposium on Remote Sensing, Vol. 4, pp. 2069-2072, 1989.
Boissonnat, J.D., Shape, reconstruction from planar cross sections,. Computer Vision, Graphics, And Image Processing, 1988. 44 (1): p. 1-29.
Buhl, EH, Schwerdtfeger, WK and P Germroth, Intracellular injection of neurons in fixed brain tissue combined with other neuroanatomical techniques at the light and electron microscopic level, In: Bjorklund, A, T Hokfelt, FG Wouterlood and AN van den Pol eds., Handbook of Chemical Neuroanatomy vol:8 ch 5 pg 273-304, 1990.
Cajal, SR, Quelques methodes de coloration des cylindres, axes, des neurofibrilles et des nids nerveux. Trav. Lab. Rech. Biol. 3:1-7, 1904.
Callaghan, P. T. (1991). Principles of Nuclear Magnetic Resonance Microscopy. New York, Oxford University Press.
Carr, H. Y. and E. M. Purcell, Effects of Diffusion on Free Precession in Nuclear Magnetic Experiments, Phys. Rev. 94, 630 (1954).
Cho, Z. H., C. B. Ahn, S. C. Juh, H. K. Lee, R. E. Jacobs, S. Lee, J. H. Yi and J. M. Jo. (1988). NMR microscopy with 4-um resolution theoretical study and experimental results. Med Phys. 15(6): 815-824.
Christian, H.N. and T.W. Sederberg, Conversion of complex contour line definitions into polygonal element mosaics. Computer Graphics, 1978. 12 (3): p. 187-192.
Christiansen, H.N. and T.W. Sederberg, Conversion of Complex Contour Line Definitions Into Polygonal Element Mosaics. Computer Graphics, 1978. 12: p. 187-192.
Chung, Y. L., A. Williams, J. S. Beech, S. C. R. Williams, J. D. Bell, I. J. Cox and J. Hope. (1995). MRI assessment of the blood-brain-barrier in a hamster model of scrapie. Neurodegeneration. 4(2): 203-207.
Clark, R.N., V.V. Trude, Cathy Ager and Gregg Swayze, Initial Vegetation Species and Senescence/Stress Indicator Mapping in the San Luis Valley, Colorado using Imaging Spectrometer Data, 5th Annual JPL Airborne Earth Science Workshop, JPL 95-1, vol. 1, pp. 35-38, 1995.
Clark, RN and T. L. Roush, Reflectance Spectroscopy--Quantitative Analysis Techniques for Remote Sensing Applications, Journal of Geophysical Research 89, pp. 6329-6340, 1984.
Clark VP, E Courchesne and M Grafe, In vivo myeloarchitectonic analysis of human striate and extrastriate cortex using magnetic resonance imaging, Cerebral Cortex, 1992, 2(5):417-24.
Conturo, T. E. et al., Encoding of Anisotropic Diffusion with Tetragonal Gradients: A General Mathematical Diffusion Formalism and Experimental Results, Magn. Reson. Med. 35, 399 (1996).
Doran, M. et al., Normal and Abnormal White Matter Tracts Shown by MR Imaging using Directional Diffusion Weighted Sequences, J. Comput. Assist. Tomogr. 14, 865 (1990).
Duvernoy, H. M. (1988) The human hippocampus. Bergmann Verlag, Munich.
Ekoule, A.B., F. Peyrin, and C. Odet (1991) A triangulation algorithm from arbitrary shaped multiple planar contours. Trans Graph 10(2):182-199.
Ekstrom P. Anterograde and retrograde filling of central neuronal systems with horseradish peroxidase under in vitro conditions. Journal of Neuroscience Methods, 1985 Oct, 15(1):21-35.
Fink, S. (1992) A solvent-free coating-procedure for the improved preparation of cryostat sections in light microscope histochemistry. Histochemistry 97:243-246.
Fuchs, H., Z.M. Kedem, and S.P. Uselton, Optimal Surface Reconstruction from Planar Contours. Communications of the ACM, 1977. 20(No.10): p. 693-702.
Fuchs, H.e.a. A System For Automatic Acquisition of Three-Dimensional Data. in Proceedings of the National Computer Conference. 1977.
Ganapathy, S., and T. Dennehy (1991) A new general triangulation algorithm for planar contours. Trans Graph 10(2):182-199.
Ghosh, P. et al., Pure Phase-Encoded MRI and Classification of Solids, IEEE Trans. Med. Imag. 14, 616 (1995).
Goverman, J. et al., Transgenic Mice that Express a Myelin Basic Protein-Specific T Cell Receptor Develop Spontaneous Autoimmunity, Cell 72, 551 (1993).
Gulani, V. et al., Rapid Diffusion Tensor Microimaging of Excised Rat Spinal Cords, Proc. Int. Soc. Magn. Res. Med., New York, 1321 (1996).
Haase, A., Snapshot Flash MRI. Applications to T1, T2 and Chemical Shift Imaging, Magn. Res. Med. 13, 77 (1990).
Haber S. Tracing intrinsic fiber connections in postmortem human brain with WGA-HRP. Journal of Neuroscience Methods, 1988 Feb, 23(1):15-22.
Hapke, B., Bi-directional Reflectance Spectroscopy, Journal of Geophysical Research 86 B4, pp. 3039-3064, 1984.
Heimer, L and Robards, J (eds), Neuroanatomical tract-tracing methods, Plenum Press, New York, 1981.
Heinsen, and Y. L. Heinsen (1991) Serial thick, frozen, gallocyanin stained sections of human central nervous system. J. Histotech 14:167-169.
Hendrickson, A and Edwards, SB, The use of axonal transport for autoradiographic tracing of pathways in the central nervous system. In: RT Robertson ed. Neuroanatomical Research Techniques, vol 2 ch 9 pg 242-291, Academic Press, New York, 1978.
Hibbard, L. and R. Hawkins (1988) Objective image alignment for three-dimensional reconstruction of digital autoradiograms J Neurosci Meth 26:55-74.
Hill, E. L., and R. Elde (1990) An improved method for preparing cryostat sections of undecalcified bone for multipe uses. J. Histochem 38:443-448.
Hine, B., and R. Rodriguez (1992) Rapid gelatin embedding procedure for frozen brain tissue sectioning. J. Histotech 15:121-122.
[pic]
Holmbom, B., M. Lindstrom, U. Naslund, and L.-E. Thornell (1991) A method for enzyme- and immunohistochemical staining of large frozen specimens. Histochemistry 95:441-447.
Holmes C. J, Bearman G., Faust J., Biswas A. and Toga A., A Paler Shade of White: Multispectral Tissue Classification of Blockface Images During Human Brain Cryosectioning, Third Annual Conference on Functional Mapping of the Human Brain, 1997.
Holmes C. J., Mainville L. S. and Jones B. E., Distribution of cholinergic, GABAergic and serotonergic neurons in the medial medullary reticular formation and their projections studied by cytotoxic lesions in the cat, Neuroscience 62(4):1155-1178, 1994.
Holmes, CJ, R Hoge, L Collins, R Woods, A Toga and A Evans, Enhancement of Magnetic Resonance Images Using Registration for Signal Averaging, submitted.
Holz, M., H. Weingartner, Calibration in Accurate Spin-Echo Diffusion Measurements Using 1H and Less-Common Nuclei, J. Magn. Reson. 92, 115 (1991).
Honig, M. G., and Hume, R. I. (1986). Fluorescent carbocyanine dyes allow living neurons of identical origin to be studied in long-term cultures. Journal of Cell Biology, 103, 171-87.
House, W. V. (1984). NMR Microscopy. IEEE Trans. Nucl. Sci. NS-31: 570-577.
Huang, W., I. Palyka, H. F. Li, E. M. Eisenstein, N. D. Volkow and C. S. Springer. (1996). Magnetic resonance imaging (MRI) detection of the murine brain response to light - temporal differentiation and negative functional MRI changes. Proceedings of The National Academy of Sciences USA. 93(12): 6037-6042.
Jacobs, R.E., Ahrens, E.T., Dickinson, M.E. & Laidlaw, D. Towards a MicroMRI Atlas of Mouse Development. Computerized Medical Imaging and Graphics 23, 15-24 (1999)
Jackels, S. C. (1990). Enhancement Agents in Magnetic Resonance and Ultrasound. Pharm. Med. Imag. Sect III Chapt 20 p645.
Jansen AS; Loewy AD. Viral tracing of innervation. Science, 1994 Jul 1, 265(5168):121-2.
Joseph, P. M. and D. Lu. (1989). A Technique for Double Resonant Operation of Birdcage Imaging Coils. IEEE Trans. Med. Imaging. 8(3): 286-294.
Jossan, S.S., P. Gillberg, R. D, Argy, S. AquiUCLAus, B. Langstrom, C. Halldin, and L. Oreland (1991) Quantitative localization of human brain monoamine oxidase B by large section autoradiography using L-[3H]deprenyl. Brain Research 547:60-76.
Jouk, P S; Usson, Y; Michalowicz, G; Grossi, L; Parazza, F. Orientation of myocardial fibres of the ventricular mass during fetal development. Mapping by means of polarized light and multimodal image analysis. (Third Conference of the European Society for Analytical Cellular Pathology, v.6, n.3, (1994): 258.
Kageyama GH; Meyer RL. Dense HRP filling in pre-fixed brain tissue for light and electron microscopy. Journal of Histochemistry and Cytochemistry, 1987 Oct, 35(10):1127-36.
Kennedy, D. N., C. Chang, V. S. Caviness, J. Moore, T. J. Brady and B. R. Rosen. (1989). Neuroanatomic Microscopy of the Rodent Brain: Imaging Strategies and Contrast. Abstracts of the Society of Magnetic Resonance in Medicine, 8th Annual Meeting, 1989: 979.
Keppel, E., Approximating Complex Surfaces by Triangulation of Contour Lines. IBM Journal of Research Development, 1975. 19: p. 2-11.
Keppel, E., Approximating complex surfaces by triangulation of contour lines, IBM. J Res. & Dev., 1975. 19 (1): p. 2-11.
Khateb A; Fort P; Alonso A; Jones BE; Muhlethaler M. Pharmacological and immunohistochemical evidence for serotonergic modulation of cholinergic nucleus basalis neurons. European Journal of Neuroscience, 1993 May 1, 5(5):541-7.
Kornguth, S., E. Bersu, M. Anderson and J. Markley. (1992). Correlation of increased levels of class-I MHC H-2K(k) in the placenta of murine trisomy-16 conceptuses with structural abnormalities revealed by magnetic resonance microscopy. Teratology. 45(4): 383-391.
Kuhn, W. (1990). NMR Microscopy - Fundamentals, Limits, and Possible Applications. Angew. Chem. Int. Engl. 29(1): 1-112.
[pic]
Kui Ying, MS, Clymer BD and Schmalbrock P., Adaptive filtering for high resolution magnetic resonance images. Journal of Magnetic Resonance Imaging, 1996 Mar-Apr, 6(2):367-77.
Laidlaw, D. K. Fleischer, and A. Barr, Classification of Material Mixtures in Volume Data for Visualization and Modeling, California Institute of Technology, Pasadena, CA, Technical Report CS-TR-94-07 (1994).
Laidlaw, D., Geometric Model Extraction from Magnetic Resonance Volume Data, Ph.D. Thesis, California Institute of Technology (1995).
Lauterbur, P. C. (1973). Image Formation by Induced Local Interactions: Examples Employing Nuclear Magnetic Resonance. Nature. 242: 190.
Le Bihan, D. et al., MR Imaging of Intravoxel Incoherent Motions: Applications to Diffusion and Perfusion in Neurologic Disorders, Radiology 161, 401 (1986).
Le Bihan, D., Molecular Diffusion Nuclear Magnetic Resonance Imaging, Magn. Reson. Q. 7, 1 (1991).
Leventhal, C., and R. Ware (1972) Three dimensional reconstruction from serial sections. Nature 236(5344):207-210.
MacKay A; Whittall K; Adler J; Li D; Paty D; Graeb D. In vivo visualization of myelin water in brain by magnetic resonance. Magnetic Resonance in Medicine, 1994 Jun, 31(6):673-7.
Maitland D. and Walsh J., Interference based linear birefringence measurements of thermal induced changes in collagen, SPIE 2134A, pp. 304-308, 1994.
Mansfield, P. and P. G. Morris. (1982). NMR Imaging in Biomedicine. New York, Academic Press.
Marko, M. A. Leith, and D. Parsons (1988) Three dimensional reconstruction of cells from serial sections and whole cell mounts using multilevel contouring of stereo micrographs. J Elec Microscope Tech 9:395-411.
Mazziotta, J.C., Toga, A.W., Evans, A., Fox, P. & Lancaster J. 1995 A Probabilistic Atlas of the Human Brain: Theory and Rationale for Its Development. NeuroImage. 2:89-101.
McDonald, J W; Roggli, V L. Detection of Silica Particles in Lung Tissue by Polarizing Light Microscopy. Archives of Pathology and Laboratory Medicine, v.119, n.3, (1995):242-246.
Mesulam, M, (Ed) Tracing neural connections with horseradish peroxidase, Wiley, New York, 1982.
Meyer, D., M. Schaefer, et al. (1990). Advances in Macrocyclic Gadolinium Complexes as Magnetic Resonance Imaging Contrast Agents. Invest. Radiol. 25: S53.
Moi, M. K. and C. F. Meares (1988). The peptide way to the Macrocyclic Bifunctional Chelating Agents, Synthesis of p-nitobenzyl DOTA. J. Am. Chem. Soc. 110: 6266.
Moseley, M. E. et al., Diffusion-Weighted MR Imaging of Anisotropic Water Diffusion in Cat Central Nervous System, Radiology 176, 353 (1990).
[pic]
Moseley, M.E., Y. Cohn, J. Kucharczyk et al., Diffusion Weighted MR Imaging of Anisotropic Water Diffusion in Cat Central Nervous System, Radiology 176, 439 (1990).
Munasinghe, J. P., G. A. Gresham, T. A. Carpenter and L. D. Hall. (1995). Magnetic Resonance Imaging of the Normal Mouse Brain: Comparison with Histologic Sections. Laboratory Animal Science. 45(6): 674-679.
Muritech, Muritech Internet Atlas of Mouse Development, 1998
Nauta, W, and LF Ryan, Selective silver impregnation of degenerating axons in the central nervous system, Staining Technology, 27:175-179, 1952.
Newton, R H; Haffegee, J P; Ho, M W. Polarized light microscopy of weakly birefringent biological specimens. Journal of Microscopy (Oxford), v.180, n.2, (1995): 127-130.
Ono, J. et al., Differentiation Between Dysmyelination and Demyelination using Magnetic Resonance Diffusional Anisotropy, Brain Res. 671, 141 (1995).
Pickering, J G; Boughner, D R. Quantification of Myocardial Collagen Using Polarized Light Microscopy and Computer-Assisted Image Analysis. Journal of Molecular and Cellular Cardiology, v.21, n.SUPPL. 2, (1989):S83.
Pickering, J G; Boughner, D R. Quantitative assessment of the age of fibrotic lesions using polarized light microscopy and digital image analysis. American Journal of Pathology, v.138, n.5, (1991): 1225-1232.
Pickering, J. G. and D. G. Boughner, Fibrosis in the transplanted heart and its relation to donor ischemic time: Assessment with polarized light microscopy and digital image analysis, Circulation, v.81, n.3, (1990): 949-958.
Press, W. H. et al., Numerical Recipes in C: The Art of Scientific Computing, (Cambridge University Press, New York, NY 1992).
Quinn, B., K. Ambach, and A.W. Toga (1993) Three-dimensional cryomacrotomy with integrated computer-based technology in neuropathology. Lab Investig (abstrt, in press)
Quinn, B., K. Ambach, and A.W. Toga, Towards a digital reconstruction human brain atlas . Soc Neurosci Abst, 1992. 18: p. :968.
Quinn, B., Myelinosomes: anovel tool for investigating basal ganglia disease. J Neuropath Exp Neurol, 1992. 51: p. 335.
Renn, O. and C. F. Meares (1992). Large scale synthesis of the bifunctional chelating agent, Synthesis of p-nitobenzyl DOTA. Bioconjugate Chem. 3: 563.
Ringwald, M., Baldock, R., Bard, J., Kaufman, M., Eppig, J.T., Richardson, J.E., Nadeau, J.H. and Davidson, D. 1994, A database for mouse development. Science 265:2033-2034.
Roland PE and Zilles K (1994). Brain atlases--a new research tool. Trends in Neurosciences, 17(11):458-67.
Rosene, D., and K. Rhodes (1990) Cryoprotection and freezing methods to control ice artifact in frozen sections of fixed and unfixed brain tissue. In Quantitative and Qualitative Microscopy, M.Conn ed., pp.360-385, Academic Press, New York.
Rouiller EM; Capt M; Dolivo M; De Ribaupierre F. Tensor tympani reflex pathways studied with retrograde horseradish peroxidase and transneuronal viral tracing techniques. Neuroscience Letters, 1986 Dec 23, 72(3):247-52.
Runge, V. M. and D. Y. Gelblumm (1991). Future Directions in Magnetic Resonance Imaging Contrast Media. Top. Magn. Reson. Imag. 3: 85.
Russell, E. J., T. F. Schaible, et al. (1989). Multicenter Double-blind Placebo-controlled Study of Gadopentetate Dimeglumine as an MR Contrast Agent: Evaluation in Patients with Cerebral Lesions. AJR 152: 813.
Cherry SR, Shao Y, Silverman RW, Chatziioannou A, Meadors K, Siegel S, Farquhar T, Young J, Jones WF, Newport D, Moyers C, Andreaco M, Paulus M, Binkley D, Nutt R, Phelps ME. MicroPET: a high resolution PET scanner for imaging small animals. IEEE Trans Nucl Sci 1997; 44: 1161-1166.
Gambhir SS, Barrio JR, Wu L, Iyer M, Namavari M, Satyamurthy N, Bauer E, Parrish C, MacLaren DC, Borghei AR, Green LA, Sharfstein A, Berk AJ, Cherry SR, Phelps ME, Herschman HR. Imaging of adenoviral-directed herpes simplex virus type 1 thymidine kinase reporter gene expression in mice with radiolabled ganciclovir. J Nucl Med 1998; 39: 2003-2011.
Gambhir SS, Barrio JR, Phelps ME, Iyer M, Namavari M, Satyamurthy N, Wu L, Green LA, Bauer E, MacLaren DC, Nguyen K, Berk AJ, Cherry SR, Herschman HR. Imaging adenoviral-directed reporter gene expression in living animals with positron emission tomography. Proc Natl Acad Sci 1999; 96: 2333-2338.
MacLaren DC, Gambhir SS, Satyamurthy N, Barrio JR, Sharfstein S, Toyokuni T, Wu L, Berk AJ, Cherry SR, Phelps ME, Herschman HR. Repetitive, non-invasive imaging of the dopamine D2 receptor as a reporter gene in living animals. Gene Therapy (in press) 1999
[pic]
Schmued, L., A rapid, sensitive histochemical stain for myelin in frozen brain sections. J Histochem Cytochem, 1990. 38: p. 717-720.
Shinagawa, Y. and T. Kunii, Constructing a (R) eeb graph automatically from cross sections. IEEE Comp Graphics and Appl, 1991. 11 (5): p. 44-51.
Shinagawa, Y., T. Kunii, and Y. Kergosien (1991) Surface coding based on (M)orse theory. IEEE Comp Graph Appl 11(5):66-78.
Shinagawa, Y., T. Kunii, and Y. Kergosien, Surface coding based on (M)orse theory. IEEE Comp Graphics and Appl, 1991. 11 (5): p. 66-78.
Stejskal, E. O.and J. E. Tanner, Spin Diffusion Measurements: Spin Echos in the Presence of a Time-Dependent Field Gradient, J. Chem. Phys. 42, 288 (1965).
Strack AM, Pseudorabies virus as a transneuronal tract tracing tool: specificity and applications to the sympathetic nervous system. Gene Therapy, 1994, 1 Suppl 1:S11-4.
Talagala, S. L. and I. J. Lowe. (1991). Introduction to Magnetic Resonance Imaging. Concepts in Magnetic Resonance. 3: 145-159.
Talairach, J. and P. Tournoux (1988) Principe et technique des etudes anatomiques. In Co-Planar Stereotaxic Atlas of the Human Brain - 3-Dimensional Proportional System: An Approach to Cerebral Imaging. M. Rayport ed., pp. 3-9, Thieme Medical Publishers, Inc., New York.
Tedeschi, C.G. (1970) Neuropathology: Methods and Diagnosis. Little, Brown and Co., Boston.
Thomsen, Sharon L.; Cheong, Wai-Fung; Pearce, John A. Changes in collagen birefringence: a quantitative histologic marker of thermal damage in skin, SPIE Vol. 1422, p. 34-42, 1991.
Toga, A. W. and T. Arnicar (1985) Image analysis of brain physiology. Comp Graph Appl 5(12):20-25.
Toga, A.W. and Banerjee, P.K. 1993 Registration revisited. J. Neurosci. Meth. 48:1-13.
Vlaverde, F. 1998, Golgi Atlas of the Postnatal Mouse Brain. Springer Verlag.
Van Gurp, M; Van Ginkel, G; Levine, Y K. Orientational properties of biological pigments in ordered systems studied with polarized light: Photosynthetic pigment-protein complexes in membranes. Journal of Theoretical Biology, v.131, n.3, (1988): 333-350.
Van Leeuwen, M. B. M, A. J. H. Deddens, P.O. Cerrits, and B. Hillen (1990) A modified mallory-cason staining procedure for large cryosections. Stain Tech 65:37-42.
van Zijl, P. C. M., D. Davis, C. T. M. Moonen, Diffusion Spectroscopy in Living Systems, in NMR in Physiology and Biomedicine, ed. R. J. Gilles, p.185 (Academic Press, New York, NY 1994).
Yagishita A; Nakano I; Oda M; Hirano A. Location of the corticospinal tract in the internal capsule at MR imaging Radiology, 1994 May, 191(2):455-60
Yan-Marriott, Y; Marriott, G. Theory, Application and comparison of real-time, fluorescence polarization image microscopy using polarized and natural-light excitation. (Thirty-fourth Annual Meeting of the American Society for Cell Biology, Molecular Biology of the Cell, v.5, n.SUPPL., (1994): 247A.
Yang, L. et al., Diffusion Contrast in NMR Microscopy of the Spinal Cord, preprint \~binglis/postertitle.html
Zaharchuk, G., H. Hara, P. L. Huang, M. A. Fishman, M. A. Moskowitz, B. G. Jenkins and B. R. Rosen. (1997). Neuronal Nitric Oxide Synthase Mutant Mice Show Smaller Infarcts and Attenuated Apparent Diffusion Coefficient Changes in the Peri-Infarct Zone During Focal Cerebral Ischemia. Magnetic Resonance in Medicine. 37(2): 170-175.
Zhou, X., and Lauterbur, P.C. (1992). NMR Microscopy Using Projection Reconstruction. In Magnetic Resonance Microscopy. (ed. B. Blumich and W. Kuhn) 3-27. New York: VCH Publishers.
1. Rugh R. The Mouse. Its Reproduction and Development. Minneapolis: Burgess Publishing Co.; reprinted 1990 by Oxford University Press, Oxford; 1968.
2. Theiler T. The House Mouse: Atlas of Embryonic Development. New York: Springer-Verlag; 1989. 178 p.
3. Kaufman MH. The Atlas of Mouse Development. London: Academic Press; 1992.
4. Williams BS, Doyle, M. An Internet atlas of mouse development. Computerized Medical Imaging and Graphics 1996;20(6):433.
5. Kaufman, M.H., Brune, R.M., Baldock, R.A., Bard, J.B.L., Davidson D. Computer-aided 3-D reconstruction of serially sectioned mouse embryos: Its use in integrating anatomical organization. International Journal of Developmental Biology 1997;41(2):223.
6. Gibaud B., Garlatti S., Barillot C., Aure E. Methodology for the design of digital brain atlases. Lecture Notes In Artificial Intelligence 1997;1211:441.
7. Toga A.W., Santori E.M., Hazani R., Ambach K. A 3d digital map of rat-brain. Brain Research Bulletin 1995;38(1):77.
8. Paxinos G., Watson C. The Rat Brain in Stereotaxic Coordinates. San Diego: Academic Press; 1997. 280 p.
9. Ghosh P., O'Dell M., Narasimhan P.T., Fraser S.E., Jacobs R.E. Mouse Lemur Microscopic MRI Brain Atlas. Neuroimage 1994;1(4):345-349 and CD-ROM.
10. Smith B.R., Linney E., Huff D.S, Johnson G.A. Magnetic Resonance Microscopy of Embryos. Computerized Medical Imaging and Graphics 1996;20(6):483-490.
11. ; ; ; ;
12. ; ; ;
13. Toga A. W., Ambach K.L., Quinn B., Shankar K., Schluender S. Postmortem Anatomy. In: Toga AW, Mazziotta JC, editors. Brain Mapping: The Methods. New York: Academic Press; 1996. p 169-190.
14. Toh M. Y., Falk R. B., Main J. S. Interactive brain atlas with the visible human project data: Development methods and techniques. Radiographics 1996;16(5):1201.
15. Narasimhan R. T., Jacobs R. E. Neuroanatomical Micormagnetic magnetic Resonance Imaging. In: Toga AW, Mazziotta JC, editors. Brain Mapping: The Methods. San Diego: Academic Press; 1996. p 147-168.
16. Jacobs R. E, Fraser S. E. Magnetic Resonance Microscopy of Embryonic Cell Lineages and Movement. Science 1994;263: 681-684.
17. Smith, B. R., Shattuck M. D. , Hedlund L. W. , Johnson G. A. Time-course imaging of rat embryos in utero with magnetic resonance microscopy. Magnetic Resonance In Medicine 1998;39(4):673.
18. Basser P.J, Pierpaoli C. Microstructural and Physiological Features of Tissues Elucidated by Quantitative-Diffusion-Tensor MRI. J. of Magnetic Resonance, Series B 1996;111:209-219.
19. Le Bihan D., editor. Diffusion and Perfusion Magnetic Resonance Imaging. New York: Raven Press; 1995.
20. Blumich B, Kuhn W, editors. Magnetic Resonance Microscopy. New York: VCH Publishers; 1992. 604 p.
21. House WV. NMR Microscopy. IEEE Trans. Nucl. Sci. 1984;NS-31:570-577.
22. Callaghan PT. Principles of Nuclear Magnetic Resonance Microscopy. New York: Oxford Unversity Press; 1991.
23. Cho, Z. H, Ahn, C. B., Juh, S. C., Jo, J. M., Friedenberg, R. M., Fraser, S. E., Jacobs, R. E.. Recent progress in nmr microscopy towards cellular imaging. Phil Trans Roy Soc London A1990;333(1632):469.
24. Zhou X, Lauterbur PC. NMR Microscopy Using Projection Reconstruction. In: Blumich b, Kuhn W, editors. Magnetic Resonance Microscopy. New York: VCH Publishers; 1992. p 3-27.
25. Cho, Z.H., Ahn, C.B., Juh, S.C., Lee, H.K., Jacobs, R.E., Lee, S., Yi, J.H., Jo, J.M. Nuclear magnetic-resonance microscopy with 4-mu-m resolution - theoretical-study and experimental results. Medical Physics 1988;15(6):815.
26. Kuhn W. NMR Microscopy - Fundementals, Limits, and Possible Applications. Angew. Chem. Int. Engl. 1990;29(1):1-112.
27. Inglis B.A, Yang L, Wirth E.D, Plant D, Marceci T.H. Diffusion Anisotropy in Excised Normal Rat Spinal Cord Measured by NMR Microscopy. Magnetic Resonance Imaging 1997;15(4):441-450.
28. Le Bihan D. Molecular diffusion, tissue microdynamics and microstructure. NMR In Biomedicine 1995;8(7-8):375.
29. Basser P.L., Le Bihan, D., Mattiello, J. Measuring tissue fiber direction using diffusion NMR. Biophysical Journal 1993;64(2):A131.
30. Moonen, C. T. W. , J. Pekar, M. H. M. de Vleeschouwer, P. van Gelderen, P. C. M. van Zijl, D. DesPres, Restricted and anisotropic displacement of water in healthy cat brain and in stroke studied by NMR diffusion imaging. Magn. Reson. Med. 19, 327-332 (1991).
31. Carr H.Y, Purcell E.M. Effects of Diffusion on Free Precession in NMR Experiments. Physical Review 1954;94:630-638.
32. Stejskal E.O., Tanner J.E. Spin Diffusion Measurements in the Presence of Time-Dependent Field Gradient. J. Chemical Physics 1965;42:288.
34. Le Bihan D., Moonen, C.T.W, Vanzijl, P.C.M., Pekar, J., Despres, D. Measuring random microscopic motion of water in tissues with mr imaging - a cat brain study. Journal of Computer Assisted Tomography 1991;15(1):19.
33. Conturo TE, McKinstry RC, Akbudak E, Robinson BH. Encoding of Anisotropic Diffusion with Tetrahedral Gradients: A General Mathematical Diffusion Formalism and Experimental Results. MRM 1996;35:399-412.
34. Ahrens, E.T., Laidlaw, D. H., Readhead, C., Brosnan, C., Fraser, S. E., & Jacobs, R. E.. Magnetic resonance microscopy of transgenic mice that spontaneously acquire Experimental Allergic Encephalomyelitis MRM 40(1), 119-132 (1998).
35. Kaufman MH, Lee KKH, Speirs S. Histological identification of primordial germ cells in diandric and digynic triploid mouse embryos. Mol. Reprod. Devel. 1990;25:364-368.
36. Moore SG. Pediatric Musculoskeletal Imaging. In: Stark DD, Bradley WG, editors. Magnetic Resonance Imaging. Volume 2. St. Louis: Mosby-Year Book; 1992. p 2223-2330.
37. Windham JP, Abd-Allah MA, Reimann DA, Froelich JW, Haggar AM. Eigenimage Filtering in MR Imaging. Journal of Computer Assisted Tomography 1988;12(1):1-9.
38. Laidlaw DH, Barr AH, Jacobs RE. Goal-Directed Magnetic Resonance Brain Micro-Imaging. In: Koslow SH, Huerta MF, editors. Neuroinformatics: An Overview of the Human Brain Projec. Mahwah, NJ: Lawrence Erlbaum Assoc; 1996. p Chapt 5.
39. Watson AD, Rocklage SM, Carvlin MJ. Contrast Agents. In: Stark DD, Bradley WG, editors. Magnetic Resonance Imaging. Volume 1. St. Louis: Mosby-Year Book, Inc.; 1992. p 372-437.
40. Runge VM, Carollo BR, Wolf CR, Nelson KL, Gelblum DY. Gd DTPA: a review of clinical indications in central nervous system magnetic resonance imaging. Radiographics 1989;9(5):929-58.
41. Jacobs R. E., Fraser S. E. Imaging Neuronal Development with Magnetic Resonance Imaging (NMR) Microscopy. In: Mize RR, Katz LC, editors. Journal of Neuroscience Methods. Volume 54. New York: Elsevier Press; 1994. p 189-196.
42. Moats R.A., Fraser S.E., Meade T.J. A ''smart'' magnetic resonance imaging agent that reports on specific enzymatic activity. Angewandte Chemie-International Edition In English 1997;36(7):726.
43. Bogdanov A, Weissleder R. The Development of in vivo imaging systems to study gene expression. TIBTECH 1998;16:5-10.
44. Wheeler DT, Schmidt R. The digital pathway: Multiphase development of a universal and expandable digital atlas of pathology. Laboratory Investigation 1997;76(1):1090.
45. Davidson D, Bard J, Brune R, Burger A, Dubreuil C, Hill W, Kaufman M, Quinn J, Stark M, Baldock R. The mouse atlas and graphical gene-expression database. Seminars In Cell & Developmental Biology 1997;8(5):509.
46. Roland PE, Zilles K. Brain Atlases - A New Research Tool. TINS 1994;17:458-467.
Lyon, MF, S Rastan, and SDM Brown. 1996. Genetic Variants and Strains of
the Laboratory Mouse.
Third ed., New York: Oxford University Press.
Papers on microPET:
Cherry SR, Shao Y, Silverman RW, Chatziioannou A, Meadors K, Siegel S, Farquhar T, Young J, Jones WF, Newport D, Moyers C, Andreaco M, Paulus M, Binkley D, Nutt R, Phelps ME. MicroPET: a high resolution PET scanner for imaging small animals. IEEE Trans Nucl Sci 1997; 44: 1161-1166.
Cherry SR, Chatziioannou A, Shao Y, Silverman RW, Meadors K, Phelps ME. Brain imaging in small animals with MicroPET. In: Quantitative Functional Brain Imaging with Positron Emission Tomography: Eds: Carson R, Daube-Witherspoon M, Herscovitch P. Academic Press, San Diego, CA, 1998 pp. 3-9.
Qi J, Leahy RM, Cherry SR, Chatziioannou A, Farquhar TH. High resolution 3D Bayesian image reconstruction using the microPET small animal scanner. Phys Med Biol 1998; 43: 1001- 1013.
Chatziioannou AF, Cherry SR, Shao Y, Silverman RW, Meadors K, Farquhar TH, Pedarsani M, Phelps ME. Performance evaluation of microPET: A high resolution LSO PET scanner for animal imaging. J Nucl Med (in press) 1999.
Papers on Imaging Gene Expression with PET:
Gambhir SS, Barrio JR, Wu L, Iyer M, Namavari M, Satyamurthy N, Bauer E, Parrish C, MacLaren DC, Borghei AR, Green LA, Sharfstein A, Berk AJ, Cherry SR, Phelps ME, Herschman HR. Imaging of adenoviral-directed herpes simplex virus type 1 thymidine kinase reporter gene expression in mice with radiolabled ganciclovir. J Nucl Med 1998; 39: 2003-2011.
Gambhir SS, Barrio JR, Phelps ME, Iyer M, Namavari M, Satyamurthy N, Wu L, Green LA, Bauer E, MacLaren DC, Nguyen K, Berk AJ, Cherry SR, Herschman HR. Imaging adenoviral-directed reporter gene expression in living animals with positron emission tomography. Proc Natl Acad Sci 1999; 96: 2333-2338.
MacLaren DC, Gambhir SS, Satyamurthy N, Barrio JR, Sharfstein S, Toyokuni T, Wu L, Berk AJ, Cherry SR, Phelps ME, Herschman HR. Repetitive, non-invasive imaging of the dopamine D2 receptor as a reporter gene in living animals. Gene Therapy (in press) 1999
APPENDIX
To facilitate the porting of programs between platforms and to present a unified interface, we have adopted several open standards for all programs. These standards are:
1) ANSI C: All code generated from the Resource is required to conform to the ANSI C standard, and to compile on any ANSI C compiler. Where use is made of platform specific functions or hardware, these functions or hardware calls must be identified by #ifdef to enable compilation to other platforms. Where functions can be easily implemented in software, optional code should be provided as workarounds for platforms lacking the specific feature.
2) OpenGL: All graphics code will use OpenGL. This industry wide standard provides a common API for graphics and will permit visually oriented programs to run across many platforms, while taking advantage of hardware support where available.
3) OpenInventor/VRML2.0: All polygon based file formats and viewing tools will use OpenInventor compatible formats, which also conform to the VRML2.0 standard. This will facilitate inter-site exchange of data as well as web-based dissemination of data and results.
4) MINC: The common image file format that has been adopted is the Medical Image –NetCDF file format. Use of a common file format greatly facilitates the interoperability of processing tools and the exchange of data between centers. All extensions for multi-dimensional display and analysis will be relative to the MINC file format, reducing the likelihood of redundancies
Quality Assurance
We evaluate the quality of software according to the following criteria: Accuracy, Robustness, Efficiency and Usability.
Accuracy. With each change in software comes the possibility of introducing new errors into the programs. We check the consistency and accuracy of computational tools by comparing their output with previously calculated and validated results. If the variance between the program’s results and a “gold standard” exceeds an error threshold, the program is deemed unready for use. For complex tools, we test all significant subsystems and interim calculations, as well as final results.
Robustness. Unstable software is, in our experience, unusable. Robust software gracefully terminates when problems are encountered and takes pains to inform the user of the conditions that caused the difficulty. Our programs do this through diligent attention to the details of software engineering.
Specific examples of this are: Always checking the status of system resource requests (such as memory allocation), Testing for “reasonableness” of interim results (e.g. fail on divides by zero), extensive use of “preflight check” software engineering techniques, and writing error messages which are meaningful to the users.
Efficiency. The value of an program is strongly related to it’s computational efficiency. This is a particular concern with our large image datasets. We address this by several means: First, we strive to create fast algorithms. Extensive run-time performance profiling is done to identify and eliminate processing bottlenecks. Where possible, we write code that will take advantage of multiple processors. This is usually done by structuring the internal architecture of our programs to sensibly map onto multi-threaded processes and by using batch queue systems to perform large scale load balancing on the various machines here. The mixture of large (program) and small (program thread) granularity enables us to easily and flexibly get the best from our computing resources.
Usability. Useful software must be understandable and easily operated. We accomplish this by involving representatives of our user community during the various phases of software design and implementation. We prototype our programs with computer novices and with experts. Particular importance is given to “no assistance” tests – software implementers passively observe and record the efforts of novices when using their programs. Program documentation is required to be intelligible to the average user, and should include “real world” examples of how the tools are used.
Software Portability
It is difficult to write portable software for the UNIX family of computers. Our experience in developing for several major vendors (SUN, SGI, IBM, DEC) has led us to software development guidelines which allow our code to compile and run on a wide variety of architectures. We use Imake and GNU Configure to compile and install our programs. Internal to the code, we make extensive use of conditional compilation directives to isolate platform and vendor specific features, include files, and library names.
Software Distribution
We use a simple alpha, beta and final release schedule. We try to make a public release once every 6 months.
1. Approximately a month before the final release date, announce to all internal developers a freeze on updates. This constitutes the alpha release. From this point until final release, only critical bug fixes are allowed. No new features are permitted.
2. Review the status of all tests. Run the tests on several different Unix platforms. Resolve all discrepancies. Repair all source code to remove compiler warnings.
3. Run quality assurance procedure. Repair all errors and warnings reported.
4. Setup e-mail mediated discussion lists to co-ordinate public release and ensure effective communications between the user community and the developers.
5. Prepare a beta release. At this point, tag the program with the beta release number, e.g. 2.0 beta. From this point until the final release, the release source code directory will be updated manually. No general updates will be performed.
6. Release the beta version to the user community. Ask for rapid feedback and fix any errors as quickly as possible. Several betas may be required.
7. Prepare a final release.
8. Compare the current release with the previous release. Prepare a document that describes the changes.
9. Package the release. Use gzip and tar for Unix releases.
10. Announce the final release to the user community.
Programmer Guidelines
We have created software guidelines that contain coding, naming and documentation standards. Adherence to standards:
1. Makes it easier for developers to understand other developers’ code.
2. Makes it easier for users to learn a system.
3. Makes it possible to generate documentation, web pages and testing scripts.
User Community Support
We make extensive use of the Internet to provide software support to the user community.
Web Site. We maintain a web site on the external Internet that points users to software resources such as releases and documentation. These are the main way of supporting our programs. Our software web sites include lists of Frequently Asked Questions (FAQ’s), progress reports on the software, developer notes, application notes, and examples of how and why the software is used.
Electronic Mailing Lists. We also maintain a mailing list for communications pertinent to the software. These are administered with the popular “Majordomo” list server software. This software automatically performs most of the tasks of running an e-mail discussion group.
HTML Man Pages
We automatically create html man pages by processing the embedded documentation that resides in our C++ code. An index is automatically created.
Application Notes. We generate html pages of application notes. These typically present a case study of how to use the software to accomplish a specific class of problems. For each case study we provide a brief explanation of the problem and its solution with the tools. We also provide sample code (or scripts) and data that can be used to repeat the case study on the reader’s computer.
-----------------------
Figure 4. A cryosectioning apparatus (left) comprises a –20o C freezer in which a powered sled carries a frozen specimen under a microtome knife. One of either a commercial RGB CCD camera (shown) or a grayscale CCD camera equipped with a tunable liquid crystal filter (not shown) is installed with the plane of focus held constant with respect to the block face. As each section is cut from the block, either a single RGB image or a sequence of grayscale images at varying frequencies (~200 to 1000 nm at as low as 5nm intervals) is captured. This data forms a "spectral cube" for each slice imaged.
CHANGE FIGS TO MOUSE SLICES
[pic]
[pic]
[pic]
................
................
In order to avoid copyright disputes, this page is only a partial summary.
To fulfill the demand for quickly locating and searching documents.
It is intelligent file search solution for home and business.