Static.cambridge.org



Supplemental InformationParticipantsDataset 1 was obtained from the Neuroimaging Informatics Tools and Resources Clearinghouse (NITRC) website and was provided by the Centers of Biomedical Research Excellence (COBRE; cobre.html). In this dataset, a diagnosis of schizophrenia was made using the Structured Clinical Interview for DSM Disorders (SCID; Diagnostic and Statistical Manual of Mental Disorders, DSM-IV) ADDIN EN.CITE <EndNote><Cite><Author>First</Author><Year>2012</Year><RecNum>1780</RecNum><DisplayText>(First<style face="italic"> et al.</style>, 2012)</DisplayText><record><rec-number>1780</rec-number><foreign-keys><key app="EN" db-id="few0ewrtoapwveerppypr29sf20p2rdezdp5" timestamp="1495144001">1780</key></foreign-keys><ref-type name="Book">6</ref-type><contributors><authors><author>First, M. B</author><author>Spitzer, R. L</author><author>Gibbon, M</author><author>Williams, J. B. W</author><author>Spitzer, R</author><author>Gibbons, M</author><author>Williams, J</author></authors></contributors><titles><title>Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I/P)</title></titles><dates><year>2012</year></dates><publisher>John Wiley &amp; Sons, Inc.</publisher><urls></urls></record></Cite></EndNote>(First et al., 2012). Exclusion criteria included confirmed or suspected pregnancy, any history of neurological disorders and a history of intellectual disability. The patients were all receiving various antipsychotic medications at the time of the study (no medication changes in 1 month). Written informed consent was obtained from participants according to institutional guidelines at the University of New Mexico.Dataset 2, acquired as part of the UCLA Consortium for Neuropsychiatric Phenomics LA5c Study, was obtained from the OpenfMRI database (accession number: ds000030). All patients underwent a semi-structured assessment with the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) ADDIN EN.CITE <EndNote><Cite><Author>First</Author><Year>2004</Year><RecNum>2672</RecNum><DisplayText>(First<style face="italic"> et al.</style>, 2004)</DisplayText><record><rec-number>2672</rec-number><foreign-keys><key app="EN" db-id="few0ewrtoapwveerppypr29sf20p2rdezdp5" timestamp="1510427175">2672</key></foreign-keys><ref-type name="Book">6</ref-type><contributors><authors><author>First, Michael B</author><author>Frances, Allen</author><author>Pincus, Harold Alan</author></authors></contributors><titles><title>DSM-IV-TR guidebook</title></titles><dates><year>2004</year></dates><publisher>American Psychiatric Pub.</publisher><urls></urls></record></Cite></EndNote>(First et al., 2004). Exclusion criteria included left-handedness, pregnancy, history of head injury with loss of consciousness or cognitive sequelae, or other contraindications to scanning. Stable medications were permitted for the patients and 81% of patients had at least one comorbid diagnosis PEVuZE5vdGU+PENpdGU+PEF1dGhvcj5Qb2xkcmFjazwvQXV0aG9yPjxZZWFyPjIwMTY8L1llYXI+

PFJlY051bT4yMTg8L1JlY051bT48RGlzcGxheVRleHQ+KFBvbGRyYWNrPHN0eWxlIGZhY2U9Iml0

YWxpYyI+IGV0IGFsLjwvc3R5bGU+LCAyMDE2KTwvRGlzcGxheVRleHQ+PHJlY29yZD48cmVjLW51

bWJlcj4yMTg8L3JlYy1udW1iZXI+PGZvcmVpZ24ta2V5cz48a2V5IGFwcD0iRU4iIGRiLWlkPSJy

ZDU1ejJldjBlZmRybWU1MmZhcDJ6djQwdnB6dGY5OTlhZGYiIHRpbWVzdGFtcD0iMTU0NzE0NjIy

MCI+MjE4PC9rZXk+PC9mb3JlaWduLWtleXM+PHJlZi10eXBlIG5hbWU9IkpvdXJuYWwgQXJ0aWNs

ZSI+MTc8L3JlZi10eXBlPjxjb250cmlidXRvcnM+PGF1dGhvcnM+PGF1dGhvcj5Qb2xkcmFjaywg

Ui4gQS48L2F1dGhvcj48YXV0aG9yPkNvbmdkb24sIEUuPC9hdXRob3I+PGF1dGhvcj5UcmlwbGV0

dCwgVy48L2F1dGhvcj48YXV0aG9yPkdvcmdvbGV3c2tpLCBLLiBKLjwvYXV0aG9yPjxhdXRob3I+

S2FybHNnb2R0LCBLLiBILjwvYXV0aG9yPjxhdXRob3I+TXVtZm9yZCwgSi4gQS48L2F1dGhvcj48

YXV0aG9yPlNhYmIsIEYuIFcuPC9hdXRob3I+PGF1dGhvcj5GcmVpbWVyLCBOLiBCLjwvYXV0aG9y

PjxhdXRob3I+TG9uZG9uLCBFLiBELjwvYXV0aG9yPjxhdXRob3I+Q2Fubm9uLCBULiBELjwvYXV0

aG9yPjxhdXRob3I+QmlsZGVyLCBSLiBNLjwvYXV0aG9yPjwvYXV0aG9ycz48L2NvbnRyaWJ1dG9y

cz48YXV0aC1hZGRyZXNzPkRlcGFydG1lbnQgb2YgUHN5Y2hvbG9neSwgU3RhbmZvcmQgVW5pdmVy

c2l0eSwgU3RhbmZvcmQsIENhbGlmb3JuaWEgOTQzMDUsIFVTQS4mI3hEO0RlcGFydG1lbnQgb2Yg

UHN5Y2hpYXRyeSBhbmQgQmlvYmVoYXZpb3JhbCBTY2llbmNlcywgVW5pdmVyc2l0eSBvZiBDYWxp

Zm9ybmlhLCBMb3MgQW5nZWxlcywgQ2FsaWZvcm5pYSA5MDA5NSwgVVNBLiYjeEQ7RGVwYXJ0bWVu

dCBvZiBQc3ljaG9sb2d5LCBVbml2ZXJzaXR5IG9mIENhbGlmb3JuaWEsIExvcyBBbmdlbGVzLCBD

YWxpZm9ybmlhIDkwMDk1LCBVU0EuJiN4RDtDZW50ZXIgZm9yIEhlYWx0aHkgTWluZHMsIFVuaXZl

cnNpdHkgb2YgV2lzY29uc2luLU1hZGlzb24sIE1hZGlzb24sIFdpc2NvbnNpbiA1MzcwNSwgVVNB

LiYjeEQ7TGV3aXMgQ2VudGVyIGZvciBOZXVyb2ltYWdpbmcsIFVuaXZlcnNpdHkgb2YgT3JlZ29u

LCBFdWdlbmUsIE9yZWdvbiA5NzQwMywgVVNBLiYjeEQ7RGVwYXJ0bWVudCBvZiBQc3ljaG9sb2d5

LCBZYWxlIFVuaXZlcnNpdHksIE5ldyBIYXZlbiwgQ29ubmVjdGljdXQgMDY1MjAsIFVTQS4mI3hE

O0RlcGFydG1lbnQgb2YgUHN5Y2hpYXRyeSwgWWFsZSBVbml2ZXJzaXR5LCBOZXcgSGF2ZW4sIENv

bm5lY3RpY3V0IDA2NTIwLCBVU0EuPC9hdXRoLWFkZHJlc3M+PHRpdGxlcz48dGl0bGU+QSBwaGVu

b21lLXdpZGUgZXhhbWluYXRpb24gb2YgbmV1cmFsIGFuZCBjb2duaXRpdmUgZnVuY3Rpb248L3Rp

dGxlPjxzZWNvbmRhcnktdGl0bGU+U2NpZW50aWZpYyBEYXRhPC9zZWNvbmRhcnktdGl0bGU+PC90

aXRsZXM+PHBlcmlvZGljYWw+PGZ1bGwtdGl0bGU+U2NpZW50aWZpYyBEYXRhPC9mdWxsLXRpdGxl

PjwvcGVyaW9kaWNhbD48cGFnZXM+MTYwMTEwPC9wYWdlcz48dm9sdW1lPjM8L3ZvbHVtZT48a2V5

d29yZHM+PGtleXdvcmQ+QWR1bHQ8L2tleXdvcmQ+PGtleXdvcmQ+QXR0ZW50aW9uIERlZmljaXQg

RGlzb3JkZXIgd2l0aCBIeXBlcmFjdGl2aXR5LypwaHlzaW9wYXRob2xvZ3k8L2tleXdvcmQ+PGtl

eXdvcmQ+Qmlwb2xhciBEaXNvcmRlci8qcGh5c2lvcGF0aG9sb2d5PC9rZXl3b3JkPjxrZXl3b3Jk

PkNvZ25pdGlvbi8qcGh5c2lvbG9neTwva2V5d29yZD48a2V5d29yZD5GZW1hbGU8L2tleXdvcmQ+

PGtleXdvcmQ+RnVuY3Rpb25hbCBOZXVyb2ltYWdpbmc8L2tleXdvcmQ+PGtleXdvcmQ+SGVhbHRo

eSBWb2x1bnRlZXJzPC9rZXl3b3JkPjxrZXl3b3JkPkh1bWFuczwva2V5d29yZD48a2V5d29yZD5J

bmZvcm1hdGlvbiBEaXNzZW1pbmF0aW9uPC9rZXl3b3JkPjxrZXl3b3JkPipJbmhpYml0aW9uIChQ

c3ljaG9sb2d5KTwva2V5d29yZD48a2V5d29yZD5NYWduZXRpYyBSZXNvbmFuY2UgSW1hZ2luZzwv

a2V5d29yZD48a2V5d29yZD5NYWxlPC9rZXl3b3JkPjxrZXl3b3JkPk1lbW9yeS8qcGh5c2lvbG9n

eTwva2V5d29yZD48a2V5d29yZD5NaWRkbGUgQWdlZDwva2V5d29yZD48a2V5d29yZD5TY2hpem9w

aHJlbmlhLypwaHlzaW9wYXRob2xvZ3k8L2tleXdvcmQ+PGtleXdvcmQ+VGFzayBQZXJmb3JtYW5j

ZSBhbmQgQW5hbHlzaXM8L2tleXdvcmQ+PGtleXdvcmQ+WW91bmcgQWR1bHQ8L2tleXdvcmQ+PC9r

ZXl3b3Jkcz48ZGF0ZXM+PHllYXI+MjAxNjwveWVhcj48cHViLWRhdGVzPjxkYXRlPkRlYyA2PC9k

YXRlPjwvcHViLWRhdGVzPjwvZGF0ZXM+PGlzYm4+MjA1Mi00NDYzIChFbGVjdHJvbmljKSYjeEQ7

MjA1Mi00NDYzIChMaW5raW5nKTwvaXNibj48YWNjZXNzaW9uLW51bT4yNzkyMjYzMjwvYWNjZXNz

aW9uLW51bT48dXJscz48cmVsYXRlZC11cmxzPjx1cmw+aHR0cDovL3d3dy5uY2JpLm5sbS5uaWgu

Z292L3B1Ym1lZC8yNzkyMjYzMjwvdXJsPjwvcmVsYXRlZC11cmxzPjwvdXJscz48Y3VzdG9tMj5Q

TUM1MTM5NjcyPC9jdXN0b20yPjxlbGVjdHJvbmljLXJlc291cmNlLW51bT4xMC4xMDM4L3NkYXRh

LjIwMTYuMTEwPC9lbGVjdHJvbmljLXJlc291cmNlLW51bT48L3JlY29yZD48L0NpdGU+PC9FbmRO

b3RlPn==

ADDIN EN.CITE PEVuZE5vdGU+PENpdGU+PEF1dGhvcj5Qb2xkcmFjazwvQXV0aG9yPjxZZWFyPjIwMTY8L1llYXI+

PFJlY051bT4yMTg8L1JlY051bT48RGlzcGxheVRleHQ+KFBvbGRyYWNrPHN0eWxlIGZhY2U9Iml0

YWxpYyI+IGV0IGFsLjwvc3R5bGU+LCAyMDE2KTwvRGlzcGxheVRleHQ+PHJlY29yZD48cmVjLW51

bWJlcj4yMTg8L3JlYy1udW1iZXI+PGZvcmVpZ24ta2V5cz48a2V5IGFwcD0iRU4iIGRiLWlkPSJy

ZDU1ejJldjBlZmRybWU1MmZhcDJ6djQwdnB6dGY5OTlhZGYiIHRpbWVzdGFtcD0iMTU0NzE0NjIy

MCI+MjE4PC9rZXk+PC9mb3JlaWduLWtleXM+PHJlZi10eXBlIG5hbWU9IkpvdXJuYWwgQXJ0aWNs

ZSI+MTc8L3JlZi10eXBlPjxjb250cmlidXRvcnM+PGF1dGhvcnM+PGF1dGhvcj5Qb2xkcmFjaywg

Ui4gQS48L2F1dGhvcj48YXV0aG9yPkNvbmdkb24sIEUuPC9hdXRob3I+PGF1dGhvcj5UcmlwbGV0

dCwgVy48L2F1dGhvcj48YXV0aG9yPkdvcmdvbGV3c2tpLCBLLiBKLjwvYXV0aG9yPjxhdXRob3I+

S2FybHNnb2R0LCBLLiBILjwvYXV0aG9yPjxhdXRob3I+TXVtZm9yZCwgSi4gQS48L2F1dGhvcj48

YXV0aG9yPlNhYmIsIEYuIFcuPC9hdXRob3I+PGF1dGhvcj5GcmVpbWVyLCBOLiBCLjwvYXV0aG9y

PjxhdXRob3I+TG9uZG9uLCBFLiBELjwvYXV0aG9yPjxhdXRob3I+Q2Fubm9uLCBULiBELjwvYXV0

aG9yPjxhdXRob3I+QmlsZGVyLCBSLiBNLjwvYXV0aG9yPjwvYXV0aG9ycz48L2NvbnRyaWJ1dG9y

cz48YXV0aC1hZGRyZXNzPkRlcGFydG1lbnQgb2YgUHN5Y2hvbG9neSwgU3RhbmZvcmQgVW5pdmVy

c2l0eSwgU3RhbmZvcmQsIENhbGlmb3JuaWEgOTQzMDUsIFVTQS4mI3hEO0RlcGFydG1lbnQgb2Yg

UHN5Y2hpYXRyeSBhbmQgQmlvYmVoYXZpb3JhbCBTY2llbmNlcywgVW5pdmVyc2l0eSBvZiBDYWxp

Zm9ybmlhLCBMb3MgQW5nZWxlcywgQ2FsaWZvcm5pYSA5MDA5NSwgVVNBLiYjeEQ7RGVwYXJ0bWVu

dCBvZiBQc3ljaG9sb2d5LCBVbml2ZXJzaXR5IG9mIENhbGlmb3JuaWEsIExvcyBBbmdlbGVzLCBD

YWxpZm9ybmlhIDkwMDk1LCBVU0EuJiN4RDtDZW50ZXIgZm9yIEhlYWx0aHkgTWluZHMsIFVuaXZl

cnNpdHkgb2YgV2lzY29uc2luLU1hZGlzb24sIE1hZGlzb24sIFdpc2NvbnNpbiA1MzcwNSwgVVNB

LiYjeEQ7TGV3aXMgQ2VudGVyIGZvciBOZXVyb2ltYWdpbmcsIFVuaXZlcnNpdHkgb2YgT3JlZ29u

LCBFdWdlbmUsIE9yZWdvbiA5NzQwMywgVVNBLiYjeEQ7RGVwYXJ0bWVudCBvZiBQc3ljaG9sb2d5

LCBZYWxlIFVuaXZlcnNpdHksIE5ldyBIYXZlbiwgQ29ubmVjdGljdXQgMDY1MjAsIFVTQS4mI3hE

O0RlcGFydG1lbnQgb2YgUHN5Y2hpYXRyeSwgWWFsZSBVbml2ZXJzaXR5LCBOZXcgSGF2ZW4sIENv

bm5lY3RpY3V0IDA2NTIwLCBVU0EuPC9hdXRoLWFkZHJlc3M+PHRpdGxlcz48dGl0bGU+QSBwaGVu

b21lLXdpZGUgZXhhbWluYXRpb24gb2YgbmV1cmFsIGFuZCBjb2duaXRpdmUgZnVuY3Rpb248L3Rp

dGxlPjxzZWNvbmRhcnktdGl0bGU+U2NpZW50aWZpYyBEYXRhPC9zZWNvbmRhcnktdGl0bGU+PC90

aXRsZXM+PHBlcmlvZGljYWw+PGZ1bGwtdGl0bGU+U2NpZW50aWZpYyBEYXRhPC9mdWxsLXRpdGxl

PjwvcGVyaW9kaWNhbD48cGFnZXM+MTYwMTEwPC9wYWdlcz48dm9sdW1lPjM8L3ZvbHVtZT48a2V5

d29yZHM+PGtleXdvcmQ+QWR1bHQ8L2tleXdvcmQ+PGtleXdvcmQ+QXR0ZW50aW9uIERlZmljaXQg

RGlzb3JkZXIgd2l0aCBIeXBlcmFjdGl2aXR5LypwaHlzaW9wYXRob2xvZ3k8L2tleXdvcmQ+PGtl

eXdvcmQ+Qmlwb2xhciBEaXNvcmRlci8qcGh5c2lvcGF0aG9sb2d5PC9rZXl3b3JkPjxrZXl3b3Jk

PkNvZ25pdGlvbi8qcGh5c2lvbG9neTwva2V5d29yZD48a2V5d29yZD5GZW1hbGU8L2tleXdvcmQ+

PGtleXdvcmQ+RnVuY3Rpb25hbCBOZXVyb2ltYWdpbmc8L2tleXdvcmQ+PGtleXdvcmQ+SGVhbHRo

eSBWb2x1bnRlZXJzPC9rZXl3b3JkPjxrZXl3b3JkPkh1bWFuczwva2V5d29yZD48a2V5d29yZD5J

bmZvcm1hdGlvbiBEaXNzZW1pbmF0aW9uPC9rZXl3b3JkPjxrZXl3b3JkPipJbmhpYml0aW9uIChQ

c3ljaG9sb2d5KTwva2V5d29yZD48a2V5d29yZD5NYWduZXRpYyBSZXNvbmFuY2UgSW1hZ2luZzwv

a2V5d29yZD48a2V5d29yZD5NYWxlPC9rZXl3b3JkPjxrZXl3b3JkPk1lbW9yeS8qcGh5c2lvbG9n

eTwva2V5d29yZD48a2V5d29yZD5NaWRkbGUgQWdlZDwva2V5d29yZD48a2V5d29yZD5TY2hpem9w

aHJlbmlhLypwaHlzaW9wYXRob2xvZ3k8L2tleXdvcmQ+PGtleXdvcmQ+VGFzayBQZXJmb3JtYW5j

ZSBhbmQgQW5hbHlzaXM8L2tleXdvcmQ+PGtleXdvcmQ+WW91bmcgQWR1bHQ8L2tleXdvcmQ+PC9r

ZXl3b3Jkcz48ZGF0ZXM+PHllYXI+MjAxNjwveWVhcj48cHViLWRhdGVzPjxkYXRlPkRlYyA2PC9k

YXRlPjwvcHViLWRhdGVzPjwvZGF0ZXM+PGlzYm4+MjA1Mi00NDYzIChFbGVjdHJvbmljKSYjeEQ7

MjA1Mi00NDYzIChMaW5raW5nKTwvaXNibj48YWNjZXNzaW9uLW51bT4yNzkyMjYzMjwvYWNjZXNz

aW9uLW51bT48dXJscz48cmVsYXRlZC11cmxzPjx1cmw+aHR0cDovL3d3dy5uY2JpLm5sbS5uaWgu

Z292L3B1Ym1lZC8yNzkyMjYzMjwvdXJsPjwvcmVsYXRlZC11cmxzPjwvdXJscz48Y3VzdG9tMj5Q

TUM1MTM5NjcyPC9jdXN0b20yPjxlbGVjdHJvbmljLXJlc291cmNlLW51bT4xMC4xMDM4L3NkYXRh

LjIwMTYuMTEwPC9lbGVjdHJvbmljLXJlc291cmNlLW51bT48L3JlY29yZD48L0NpdGU+PC9FbmRO

b3RlPn==

ADDIN EN.CITE.DATA (Poldrack et al., 2016). After receiving a verbal explanation of the study, participants gave written informed consent following procedures approved by the Institutional Review Boards at UCLA and the Los Angeles County Department of Mental Health.Dataset 3 was acquired at Maastricht University, The Netherlands. Patients were recruited through clinicians working in selected representative geographic areas in the Netherlands and Belgium. Diagnosis of schizophrenia was based on DSM-IV criteria ADDIN EN.CITE <EndNote><Cite><Author>Association</Author><Year>2000</Year><RecNum>1779</RecNum><DisplayText>(Association, 2000)</DisplayText><record><rec-number>1779</rec-number><foreign-keys><key app="EN" db-id="few0ewrtoapwveerppypr29sf20p2rdezdp5" timestamp="1495143469">1779</key></foreign-keys><ref-type name="Journal Article">17</ref-type><contributors><authors><author>Association, American Psychiatric</author></authors></contributors><titles><title>Diagnostic and Statistical Manual of Mental Disorders fourth edition, Text revision (DSM-IV TR)</title></titles><dates><year>2000</year></dates><urls></urls></record></Cite></EndNote>(Association, 2000), assessed with the Comprehensive Assessment of Symptoms and History (CASH) interview ADDIN EN.CITE <EndNote><Cite><Author>Andreasen</Author><Year>1992</Year><RecNum>168</RecNum><DisplayText>(Andreasen<style face="italic"> et al.</style>, 1992)</DisplayText><record><rec-number>168</rec-number><foreign-keys><key app="EN" db-id="rd55z2ev0efdrme52fap2zv40vpztf999adf" timestamp="1547146001">168</key></foreign-keys><ref-type name="Journal Article">17</ref-type><contributors><authors><author>Andreasen, N. C.</author><author>Flaum, M.</author><author>Arndt, S.</author></authors></contributors><auth-address>Mental Health Clinical Research Center, University of Iowa Hospitals and Clinics, Iowa City 52242.</auth-address><titles><title>The Comprehensive Assessment of Symptoms and History (CASH). An instrument for assessing diagnosis and psychopathology</title><secondary-title>Arch Gen Psychiatry</secondary-title></titles><periodical><full-title>Archives of General Psychiatry</full-title><abbr-1>Arch. Gen. Psychiatry</abbr-1><abbr-2>Arch Gen Psychiatry</abbr-2></periodical><pages>615-23</pages><volume>49</volume><number>8</number><keywords><keyword>Databases, Factual</keyword><keyword>Depressive Disorder/diagnosis/psychology</keyword><keyword>Follow-Up Studies</keyword><keyword>Humans</keyword><keyword>Mental Disorders/*diagnosis/psychology</keyword><keyword>Psychiatric Status Rating Scales/*instrumentation/standards/statistics &amp;</keyword><keyword>numerical data</keyword><keyword>Psychometrics</keyword><keyword>Psychotic Disorders/diagnosis/psychology</keyword><keyword>Reproducibility of Results</keyword><keyword>Schizophrenia/diagnosis</keyword><keyword>Schizophrenic Psychology</keyword><keyword>Schizotypal Personality Disorder/diagnosis/psychology</keyword></keywords><dates><year>1992</year><pub-dates><date>Aug</date></pub-dates></dates><isbn>0003-990X (Print)&#xD;0003-990X (Linking)</isbn><accession-num>1637251</accession-num><urls><related-urls><url>;(Andreasen et al., 1992). Exclusion criteria included confirmed or suspected pregnancy, any history of neurological disorders, a history of intellectual disability and/or a history of substance abuse/dependence within the last 12 months. Antipsychotic medication use for patients was determined using the reports of the participant’s psychiatrist. The ethics committee of Maastricht University approved the study, and all the participants gave written informed consent in accordance with the committee’s guidelines and the Declaration of Helsinki.Dataset 4 was acquired in Dublin and scanned at the Trinity College Institute of Neuroscience as part of a Science Foundation Ireland-funded neuroimaging genetics study ("A structural and functional MRI investigation of genetics, cognition and emotion in schizophrenia"). Patients with confirmed DSM-IV diagnosis of schizophrenia were recruited through local clinical services. Exclusion criteria included confirmed or suspected pregnancy, any history of neurological disorders or intellectual disability and substance misuse in the preceding three months. All patients were chronic, but stable, medicated outpatients, with a confirmed diagnosis. Participants provided written, informed consent in accordance with local ethics committee guidelines.Dataset 5 was acquired at the West China Hospital of Sichuan University, Chengdu, China. An initial diagnosis of schizophrenia and duration of illness were determined by consensus between 2 experienced psychiatrists, using the Structured Clinical Interview for DSM-IV(SCID)-Patient Version ADDIN EN.CITE <EndNote><Cite><Author>Association</Author><Year>2000</Year><RecNum>1779</RecNum><DisplayText>(Association, 2000)</DisplayText><record><rec-number>1779</rec-number><foreign-keys><key app="EN" db-id="few0ewrtoapwveerppypr29sf20p2rdezdp5" timestamp="1495143469">1779</key></foreign-keys><ref-type name="Journal Article">17</ref-type><contributors><authors><author>Association, American Psychiatric</author></authors></contributors><titles><title>Diagnostic and Statistical Manual of Mental Disorders fourth edition, Text revision (DSM-IV TR)</title></titles><dates><year>2000</year></dates><urls></urls></record></Cite></EndNote>(Association, 2000). In addition, diagnosis of schizophrenia was confirmed for all the patients at 1-year follow-up. Exclusion criteria were the existence of a neurological disorder or other psychiatric disorders, alcohol or drug abuse (DSM-IV), pregnancy, and any chronic physical illness such as a brain tumor, hepatitis, or epilepsy, as assessed by clinical evaluations and medical records. All patients are drug-naive first-episode schizophrenia. The study was approved by the ethics committee of West China Hospital, and written informed consent was obtained from all participants.MRI acquisitionAt each site the rs-fMRI images were acquired by the Echo-Planar Imaging (EPI) sequence. Dataset 1 was acquired using a 3T Siemens scanner. The sequence parameters were as follows: repetition time/echo time (TR/TE) = 2000/30 ms; flip angle = 90°; 33 axial slices per volume; voxel size = 3.75 × 3.75 × 4.55 mm3; number of volumes= 150. Dataset 2 was acquired using a 3T Siemens scanner. The sequence parameters were as follows: repetition time/echo time (TR/TE) = 2000/30 ms; flip angle = 90°; 34 axial slices per volume; voxel size = 3 × 3 × 4 mm3; number of volumes= 152. Dataset 3 was acquired using a 3T Siemens Magnetom Allegra head scanner. The sequence parameters were as follows: repetition time/echo time (TR/TE) = 1500/30 ms; flip angle = 90°; 27 axial slices per volume; voxel size = 3.5 × 3.5 × 5.2 mm3; number of volumes= 200. Dataset 4 was acquired using a 3T Philips Intera Achieva scanner. The sequence parameters were as follows: repetition time/echo time (TR/TE) = 2000/30 ms; flip angle = 90°; 35 axial slices per volume; voxel size = 3.5 × 3.5 × 3.5 mm3; number of volumes= 180. Dataset 5 was acquired using a 3T GE scanner (EXCITE; General Electric, Milwaukee, Wisconsin). The sequence parameters were as follows: repetition time/echo time (TR/TE) = 2000/30 ms; flip angle = 90°; 30 axial slices per volume; voxel size = 3.75 × 3.5 × 5 mm3; number of volumes= 200. Dealing with potential motion-related artifacts. To deal with potential artefacts arising from motion or other factors, we applied Friston 24-parameter correction colleagues PEVuZE5vdGU+PENpdGU+PEF1dGhvcj5ZYW48L0F1dGhvcj48WWVhcj4yMDEzPC9ZZWFyPjxSZWNO

dW0+MTEyPC9SZWNOdW0+PERpc3BsYXlUZXh0PihZYW48c3R5bGUgZmFjZT0iaXRhbGljIj4gZXQg

YWwuPC9zdHlsZT4sIDIwMTMpPC9EaXNwbGF5VGV4dD48cmVjb3JkPjxyZWMtbnVtYmVyPjExMjwv

cmVjLW51bWJlcj48Zm9yZWlnbi1rZXlzPjxrZXkgYXBwPSJFTiIgZGItaWQ9IndyeHc5enIyM3h6

emEzZTAwZHB2dnNyaHgwOWF3enB3cjBkcyIgdGltZXN0YW1wPSIxNTUwNDI0NDcxIj4xMTI8L2tl

eT48L2ZvcmVpZ24ta2V5cz48cmVmLXR5cGUgbmFtZT0iSm91cm5hbCBBcnRpY2xlIj4xNzwvcmVm

LXR5cGU+PGNvbnRyaWJ1dG9ycz48YXV0aG9ycz48YXV0aG9yPllhbiwgQy4gRy48L2F1dGhvcj48

YXV0aG9yPkNyYWRkb2NrLCBSLiBDLjwvYXV0aG9yPjxhdXRob3I+SGUsIFkuPC9hdXRob3I+PGF1

dGhvcj5NaWxoYW0sIE0uIFAuPC9hdXRob3I+PC9hdXRob3JzPjwvY29udHJpYnV0b3JzPjxhdXRo

LWFkZHJlc3M+TmF0aGFuIEtsaW5lIEluc3RpdHV0ZSBmb3IgUHN5Y2hpYXRyaWMgUmVzZWFyY2gg

T3JhbmdlYnVyZywgTlksIFVTQSA7IENlbnRlciBmb3IgdGhlIERldmVsb3BpbmcgQnJhaW4sIENo

aWxkIE1pbmQgSW5zdGl0dXRlIE5ldyBZb3JrLCBOWSwgVVNBIDsgVGhlIFBoeWxsaXMgR3JlZW4g

YW5kIFJhbmRvbHBoIENvd2VuIEluc3RpdHV0ZSBmb3IgUGVkaWF0cmljIE5ldXJvc2NpZW5jZSwg

TmV3IFlvcmsgVW5pdmVyc2l0eSBDaGlsZCBTdHVkeSBDZW50ZXIgTmV3IFlvcmssIE5ZLCBVU0Eu

JiN4RDtOYXRoYW4gS2xpbmUgSW5zdGl0dXRlIGZvciBQc3ljaGlhdHJpYyBSZXNlYXJjaCBPcmFu

Z2VidXJnLCBOWSwgVVNBIDsgQ2VudGVyIGZvciB0aGUgRGV2ZWxvcGluZyBCcmFpbiwgQ2hpbGQg

TWluZCBJbnN0aXR1dGUgTmV3IFlvcmssIE5ZLCBVU0EuJiN4RDtTdGF0ZSBLZXkgTGFib3JhdG9y

eSBvZiBDb2duaXRpdmUgTmV1cm9zY2llbmNlIGFuZCBMZWFybmluZyAmYW1wOyBJREcvTWNHb3Zl

cm4gSW5zdGl0dXRlIGZvciBCcmFpbiBSZXNlYXJjaCwgQmVpamluZyBOb3JtYWwgVW5pdmVyc2l0

eSBCZWlqaW5nLCBDaGluYSA7IENlbnRlciBmb3IgQ29sbGFib3JhdGlvbiBhbmQgSW5ub3ZhdGlv

biBpbiBCcmFpbiBhbmQgTGVhcm5pbmcgU2NpZW5jZXMsIEJlaWppbmcgTm9ybWFsIFVuaXZlcnNp

dHkgQmVpamluZywgQ2hpbmEuPC9hdXRoLWFkZHJlc3M+PHRpdGxlcz48dGl0bGU+QWRkcmVzc2lu

ZyBoZWFkIG1vdGlvbiBkZXBlbmRlbmNpZXMgZm9yIHNtYWxsLXdvcmxkIHRvcG9sb2dpZXMgaW4g

ZnVuY3Rpb25hbCBjb25uZWN0b21pY3M8L3RpdGxlPjxzZWNvbmRhcnktdGl0bGU+RnJvbnQgSHVt

IE5ldXJvc2NpPC9zZWNvbmRhcnktdGl0bGU+PC90aXRsZXM+PHBlcmlvZGljYWw+PGZ1bGwtdGl0

bGU+RnJvbnRpZXJzIGluIEh1bWFuIE5ldXJvc2NpZW5jZTwvZnVsbC10aXRsZT48YWJici0xPkZy

b250LiBIdW0uIE5ldXJvc2NpLjwvYWJici0xPjxhYmJyLTI+RnJvbnQgSHVtIE5ldXJvc2NpPC9h

YmJyLTI+PC9wZXJpb2RpY2FsPjxwYWdlcz45MTA8L3BhZ2VzPjx2b2x1bWU+Nzwvdm9sdW1lPjxr

ZXl3b3Jkcz48a2V5d29yZD5mdW5jdGlvbmFsIGNvbm5lY3RvbWljczwva2V5d29yZD48a2V5d29y

ZD5oZWFkIG1vdGlvbiBpbXBhY3Q8L2tleXdvcmQ+PGtleXdvcmQ+bmV0d29yayBhbmFseXNpczwv

a2V5d29yZD48a2V5d29yZD5yZXN0aW5nLXN0YXRlIGZNUkk8L2tleXdvcmQ+PGtleXdvcmQ+c21h

bGwtd29ybGQ8L2tleXdvcmQ+PGtleXdvcmQ+dG9wb2xvZ2ljYWwgcGFyYW1ldGVyczwva2V5d29y

ZD48L2tleXdvcmRzPjxkYXRlcz48eWVhcj4yMDEzPC95ZWFyPjwvZGF0ZXM+PGlzYm4+MTY2Mi01

MTYxIChQcmludCkmI3hEOzE2NjItNTE2MSAoTGlua2luZyk8L2lzYm4+PGFjY2Vzc2lvbi1udW0+

MjQ0MjE3NjQ8L2FjY2Vzc2lvbi1udW0+PHVybHM+PHJlbGF0ZWQtdXJscz48dXJsPmh0dHA6Ly93

d3cubmNiaS5ubG0ubmloLmdvdi9wdWJtZWQvMjQ0MjE3NjQ8L3VybD48L3JlbGF0ZWQtdXJscz48

L3VybHM+PGN1c3RvbTI+UE1DMzg3MjcyODwvY3VzdG9tMj48ZWxlY3Ryb25pYy1yZXNvdXJjZS1u

dW0+MTAuMzM4OS9mbmh1bS4yMDEzLjAwOTEwPC9lbGVjdHJvbmljLXJlc291cmNlLW51bT48L3Jl

Y29yZD48L0NpdGU+PC9FbmROb3RlPn==

ADDIN EN.CITE PEVuZE5vdGU+PENpdGU+PEF1dGhvcj5ZYW48L0F1dGhvcj48WWVhcj4yMDEzPC9ZZWFyPjxSZWNO

dW0+MTEyPC9SZWNOdW0+PERpc3BsYXlUZXh0PihZYW48c3R5bGUgZmFjZT0iaXRhbGljIj4gZXQg

YWwuPC9zdHlsZT4sIDIwMTMpPC9EaXNwbGF5VGV4dD48cmVjb3JkPjxyZWMtbnVtYmVyPjExMjwv

cmVjLW51bWJlcj48Zm9yZWlnbi1rZXlzPjxrZXkgYXBwPSJFTiIgZGItaWQ9IndyeHc5enIyM3h6

emEzZTAwZHB2dnNyaHgwOWF3enB3cjBkcyIgdGltZXN0YW1wPSIxNTUwNDI0NDcxIj4xMTI8L2tl

eT48L2ZvcmVpZ24ta2V5cz48cmVmLXR5cGUgbmFtZT0iSm91cm5hbCBBcnRpY2xlIj4xNzwvcmVm

LXR5cGU+PGNvbnRyaWJ1dG9ycz48YXV0aG9ycz48YXV0aG9yPllhbiwgQy4gRy48L2F1dGhvcj48

YXV0aG9yPkNyYWRkb2NrLCBSLiBDLjwvYXV0aG9yPjxhdXRob3I+SGUsIFkuPC9hdXRob3I+PGF1

dGhvcj5NaWxoYW0sIE0uIFAuPC9hdXRob3I+PC9hdXRob3JzPjwvY29udHJpYnV0b3JzPjxhdXRo

LWFkZHJlc3M+TmF0aGFuIEtsaW5lIEluc3RpdHV0ZSBmb3IgUHN5Y2hpYXRyaWMgUmVzZWFyY2gg

T3JhbmdlYnVyZywgTlksIFVTQSA7IENlbnRlciBmb3IgdGhlIERldmVsb3BpbmcgQnJhaW4sIENo

aWxkIE1pbmQgSW5zdGl0dXRlIE5ldyBZb3JrLCBOWSwgVVNBIDsgVGhlIFBoeWxsaXMgR3JlZW4g

YW5kIFJhbmRvbHBoIENvd2VuIEluc3RpdHV0ZSBmb3IgUGVkaWF0cmljIE5ldXJvc2NpZW5jZSwg

TmV3IFlvcmsgVW5pdmVyc2l0eSBDaGlsZCBTdHVkeSBDZW50ZXIgTmV3IFlvcmssIE5ZLCBVU0Eu

JiN4RDtOYXRoYW4gS2xpbmUgSW5zdGl0dXRlIGZvciBQc3ljaGlhdHJpYyBSZXNlYXJjaCBPcmFu

Z2VidXJnLCBOWSwgVVNBIDsgQ2VudGVyIGZvciB0aGUgRGV2ZWxvcGluZyBCcmFpbiwgQ2hpbGQg

TWluZCBJbnN0aXR1dGUgTmV3IFlvcmssIE5ZLCBVU0EuJiN4RDtTdGF0ZSBLZXkgTGFib3JhdG9y

eSBvZiBDb2duaXRpdmUgTmV1cm9zY2llbmNlIGFuZCBMZWFybmluZyAmYW1wOyBJREcvTWNHb3Zl

cm4gSW5zdGl0dXRlIGZvciBCcmFpbiBSZXNlYXJjaCwgQmVpamluZyBOb3JtYWwgVW5pdmVyc2l0

eSBCZWlqaW5nLCBDaGluYSA7IENlbnRlciBmb3IgQ29sbGFib3JhdGlvbiBhbmQgSW5ub3ZhdGlv

biBpbiBCcmFpbiBhbmQgTGVhcm5pbmcgU2NpZW5jZXMsIEJlaWppbmcgTm9ybWFsIFVuaXZlcnNp

dHkgQmVpamluZywgQ2hpbmEuPC9hdXRoLWFkZHJlc3M+PHRpdGxlcz48dGl0bGU+QWRkcmVzc2lu

ZyBoZWFkIG1vdGlvbiBkZXBlbmRlbmNpZXMgZm9yIHNtYWxsLXdvcmxkIHRvcG9sb2dpZXMgaW4g

ZnVuY3Rpb25hbCBjb25uZWN0b21pY3M8L3RpdGxlPjxzZWNvbmRhcnktdGl0bGU+RnJvbnQgSHVt

IE5ldXJvc2NpPC9zZWNvbmRhcnktdGl0bGU+PC90aXRsZXM+PHBlcmlvZGljYWw+PGZ1bGwtdGl0

bGU+RnJvbnRpZXJzIGluIEh1bWFuIE5ldXJvc2NpZW5jZTwvZnVsbC10aXRsZT48YWJici0xPkZy

b250LiBIdW0uIE5ldXJvc2NpLjwvYWJici0xPjxhYmJyLTI+RnJvbnQgSHVtIE5ldXJvc2NpPC9h

YmJyLTI+PC9wZXJpb2RpY2FsPjxwYWdlcz45MTA8L3BhZ2VzPjx2b2x1bWU+Nzwvdm9sdW1lPjxr

ZXl3b3Jkcz48a2V5d29yZD5mdW5jdGlvbmFsIGNvbm5lY3RvbWljczwva2V5d29yZD48a2V5d29y

ZD5oZWFkIG1vdGlvbiBpbXBhY3Q8L2tleXdvcmQ+PGtleXdvcmQ+bmV0d29yayBhbmFseXNpczwv

a2V5d29yZD48a2V5d29yZD5yZXN0aW5nLXN0YXRlIGZNUkk8L2tleXdvcmQ+PGtleXdvcmQ+c21h

bGwtd29ybGQ8L2tleXdvcmQ+PGtleXdvcmQ+dG9wb2xvZ2ljYWwgcGFyYW1ldGVyczwva2V5d29y

ZD48L2tleXdvcmRzPjxkYXRlcz48eWVhcj4yMDEzPC95ZWFyPjwvZGF0ZXM+PGlzYm4+MTY2Mi01

MTYxIChQcmludCkmI3hEOzE2NjItNTE2MSAoTGlua2luZyk8L2lzYm4+PGFjY2Vzc2lvbi1udW0+

MjQ0MjE3NjQ8L2FjY2Vzc2lvbi1udW0+PHVybHM+PHJlbGF0ZWQtdXJscz48dXJsPmh0dHA6Ly93

d3cubmNiaS5ubG0ubmloLmdvdi9wdWJtZWQvMjQ0MjE3NjQ8L3VybD48L3JlbGF0ZWQtdXJscz48

L3VybHM+PGN1c3RvbTI+UE1DMzg3MjcyODwvY3VzdG9tMj48ZWxlY3Ryb25pYy1yZXNvdXJjZS1u

dW0+MTAuMzM4OS9mbmh1bS4yMDEzLjAwOTEwPC9lbGVjdHJvbmljLXJlc291cmNlLW51bT48L3Jl

Y29yZD48L0NpdGU+PC9FbmROb3RlPn==

ADDIN EN.CITE.DATA (Yan et al., 2013). In addition, to ensure that motion artifacts were not contributing to the observed group differences, we employed the “head motion scrubbing” method proposed by Power and colleagues PEVuZE5vdGU+PENpdGU+PEF1dGhvcj5Qb3dlcjwvQXV0aG9yPjxZZWFyPjIwMTQ8L1llYXI+PFJl

Y051bT4xMTM8L1JlY051bT48RGlzcGxheVRleHQ+KFBvd2VyPHN0eWxlIGZhY2U9Iml0YWxpYyI+

IGV0IGFsLjwvc3R5bGU+LCAyMDE0KTwvRGlzcGxheVRleHQ+PHJlY29yZD48cmVjLW51bWJlcj4x

MTM8L3JlYy1udW1iZXI+PGZvcmVpZ24ta2V5cz48a2V5IGFwcD0iRU4iIGRiLWlkPSJ3cnh3OXpy

MjN4enphM2UwMGRwdnZzcmh4MDlhd3pwd3IwZHMiIHRpbWVzdGFtcD0iMTU1MDQyNDUzNCI+MTEz

PC9rZXk+PC9mb3JlaWduLWtleXM+PHJlZi10eXBlIG5hbWU9IkpvdXJuYWwgQXJ0aWNsZSI+MTc8

L3JlZi10eXBlPjxjb250cmlidXRvcnM+PGF1dGhvcnM+PGF1dGhvcj5Qb3dlciwgSi4gRC48L2F1

dGhvcj48YXV0aG9yPk1pdHJhLCBBLjwvYXV0aG9yPjxhdXRob3I+TGF1bWFubiwgVC4gTy48L2F1

dGhvcj48YXV0aG9yPlNueWRlciwgQS4gWi48L2F1dGhvcj48YXV0aG9yPlNjaGxhZ2dhciwgQi4g

TC48L2F1dGhvcj48YXV0aG9yPlBldGVyc2VuLCBTLiBFLjwvYXV0aG9yPjwvYXV0aG9ycz48L2Nv

bnRyaWJ1dG9ycz48YXV0aC1hZGRyZXNzPkRlcHQuIG9mIE5ldXJvbG9neSwgV2FzaGluZ3RvbiBV

bml2ZXJzaXR5IFNjaG9vbCBvZiBNZWRpY2luZSBpbiBTdC4gTG91aXMsIDY2MCBTLiBFdWNsaWQg

QXZlLiwgU3QuIExvdWlzLCBNTyA2MzExMCwgVVNBLiBFbGVjdHJvbmljIGFkZHJlc3M6IHBvd2Vy

akB3dXNtLnd1c3RsLmVkdS48L2F1dGgtYWRkcmVzcz48dGl0bGVzPjx0aXRsZT5NZXRob2RzIHRv

IGRldGVjdCwgY2hhcmFjdGVyaXplLCBhbmQgcmVtb3ZlIG1vdGlvbiBhcnRpZmFjdCBpbiByZXN0

aW5nIHN0YXRlIGZNUkk8L3RpdGxlPjxzZWNvbmRhcnktdGl0bGU+TmV1cm9pbWFnZTwvc2Vjb25k

YXJ5LXRpdGxlPjwvdGl0bGVzPjxwZXJpb2RpY2FsPjxmdWxsLXRpdGxlPk5ldXJvaW1hZ2U8L2Z1

bGwtdGl0bGU+PGFiYnItMT5OZXVyb2ltYWdlPC9hYmJyLTE+PGFiYnItMj5OZXVyb2ltYWdlPC9h

YmJyLTI+PC9wZXJpb2RpY2FsPjxwYWdlcz4zMjAtNDE8L3BhZ2VzPjx2b2x1bWU+ODQ8L3ZvbHVt

ZT48a2V5d29yZHM+PGtleXdvcmQ+QWRvbGVzY2VudDwva2V5d29yZD48a2V5d29yZD5BZHVsdDwv

a2V5d29yZD48a2V5d29yZD5BbGdvcml0aG1zPC9rZXl3b3JkPjxrZXl3b3JkPipBcnRpZmFjdHM8

L2tleXdvcmQ+PGtleXdvcmQ+QnJhaW4vKnBoeXNpb2xvZ3k8L2tleXdvcmQ+PGtleXdvcmQ+QnJh

aW4gTWFwcGluZy8qbWV0aG9kczwva2V5d29yZD48a2V5d29yZD5GZW1hbGU8L2tleXdvcmQ+PGtl

eXdvcmQ+SGVhZCBNb3ZlbWVudHMvKnBoeXNpb2xvZ3k8L2tleXdvcmQ+PGtleXdvcmQ+SHVtYW5z

PC9rZXl3b3JkPjxrZXl3b3JkPkltYWdlIEVuaGFuY2VtZW50LyptZXRob2RzPC9rZXl3b3JkPjxr

ZXl3b3JkPkltYWdlIEludGVycHJldGF0aW9uLCBDb21wdXRlci1Bc3Npc3RlZC8qbWV0aG9kczwv

a2V5d29yZD48a2V5d29yZD5NYWduZXRpYyBSZXNvbmFuY2UgSW1hZ2luZy8qbWV0aG9kczwva2V5

d29yZD48a2V5d29yZD5NYWxlPC9rZXl3b3JkPjxrZXl3b3JkPk1vdGlvbjwva2V5d29yZD48a2V5

d29yZD5QYXR0ZXJuIFJlY29nbml0aW9uLCBBdXRvbWF0ZWQvbWV0aG9kczwva2V5d29yZD48a2V5

d29yZD5SZXByb2R1Y2liaWxpdHkgb2YgUmVzdWx0czwva2V5d29yZD48a2V5d29yZD5SZXN0L3Bo

eXNpb2xvZ3k8L2tleXdvcmQ+PGtleXdvcmQ+U2Vuc2l0aXZpdHkgYW5kIFNwZWNpZmljaXR5PC9r

ZXl3b3JkPjxrZXl3b3JkPlN1YnRyYWN0aW9uIFRlY2huaXF1ZTwva2V5d29yZD48a2V5d29yZD5Z

b3VuZyBBZHVsdDwva2V5d29yZD48a2V5d29yZD5BcnRpZmFjdDwva2V5d29yZD48a2V5d29yZD5G

dW5jdGlvbmFsIGNvbm5lY3Rpdml0eTwva2V5d29yZD48a2V5d29yZD5Ncmk8L2tleXdvcmQ+PGtl

eXdvcmQ+TW92ZW1lbnQ8L2tleXdvcmQ+PGtleXdvcmQ+UmVzdGluZyBzdGF0ZTwva2V5d29yZD48

L2tleXdvcmRzPjxkYXRlcz48eWVhcj4yMDE0PC95ZWFyPjxwdWItZGF0ZXM+PGRhdGU+SmFuIDE8

L2RhdGU+PC9wdWItZGF0ZXM+PC9kYXRlcz48aXNibj4xMDk1LTk1NzIgKEVsZWN0cm9uaWMpJiN4

RDsxMDUzLTgxMTkgKExpbmtpbmcpPC9pc2JuPjxhY2Nlc3Npb24tbnVtPjIzOTk0MzE0PC9hY2Nl

c3Npb24tbnVtPjx1cmxzPjxyZWxhdGVkLXVybHM+PHVybD5odHRwOi8vd3d3Lm5jYmkubmxtLm5p

aC5nb3YvcHVibWVkLzIzOTk0MzE0PC91cmw+PC9yZWxhdGVkLXVybHM+PC91cmxzPjxjdXN0b20y

PlBNQzM4NDkzMzg8L2N1c3RvbTI+PGVsZWN0cm9uaWMtcmVzb3VyY2UtbnVtPjEwLjEwMTYvai5u

ZXVyb2ltYWdlLjIwMTMuMDguMDQ4PC9lbGVjdHJvbmljLXJlc291cmNlLW51bT48L3JlY29yZD48

L0NpdGU+PC9FbmROb3RlPn==

ADDIN EN.CITE PEVuZE5vdGU+PENpdGU+PEF1dGhvcj5Qb3dlcjwvQXV0aG9yPjxZZWFyPjIwMTQ8L1llYXI+PFJl

Y051bT4xMTM8L1JlY051bT48RGlzcGxheVRleHQ+KFBvd2VyPHN0eWxlIGZhY2U9Iml0YWxpYyI+

IGV0IGFsLjwvc3R5bGU+LCAyMDE0KTwvRGlzcGxheVRleHQ+PHJlY29yZD48cmVjLW51bWJlcj4x

MTM8L3JlYy1udW1iZXI+PGZvcmVpZ24ta2V5cz48a2V5IGFwcD0iRU4iIGRiLWlkPSJ3cnh3OXpy

MjN4enphM2UwMGRwdnZzcmh4MDlhd3pwd3IwZHMiIHRpbWVzdGFtcD0iMTU1MDQyNDUzNCI+MTEz

PC9rZXk+PC9mb3JlaWduLWtleXM+PHJlZi10eXBlIG5hbWU9IkpvdXJuYWwgQXJ0aWNsZSI+MTc8

L3JlZi10eXBlPjxjb250cmlidXRvcnM+PGF1dGhvcnM+PGF1dGhvcj5Qb3dlciwgSi4gRC48L2F1

dGhvcj48YXV0aG9yPk1pdHJhLCBBLjwvYXV0aG9yPjxhdXRob3I+TGF1bWFubiwgVC4gTy48L2F1

dGhvcj48YXV0aG9yPlNueWRlciwgQS4gWi48L2F1dGhvcj48YXV0aG9yPlNjaGxhZ2dhciwgQi4g

TC48L2F1dGhvcj48YXV0aG9yPlBldGVyc2VuLCBTLiBFLjwvYXV0aG9yPjwvYXV0aG9ycz48L2Nv

bnRyaWJ1dG9ycz48YXV0aC1hZGRyZXNzPkRlcHQuIG9mIE5ldXJvbG9neSwgV2FzaGluZ3RvbiBV

bml2ZXJzaXR5IFNjaG9vbCBvZiBNZWRpY2luZSBpbiBTdC4gTG91aXMsIDY2MCBTLiBFdWNsaWQg

QXZlLiwgU3QuIExvdWlzLCBNTyA2MzExMCwgVVNBLiBFbGVjdHJvbmljIGFkZHJlc3M6IHBvd2Vy

akB3dXNtLnd1c3RsLmVkdS48L2F1dGgtYWRkcmVzcz48dGl0bGVzPjx0aXRsZT5NZXRob2RzIHRv

IGRldGVjdCwgY2hhcmFjdGVyaXplLCBhbmQgcmVtb3ZlIG1vdGlvbiBhcnRpZmFjdCBpbiByZXN0

aW5nIHN0YXRlIGZNUkk8L3RpdGxlPjxzZWNvbmRhcnktdGl0bGU+TmV1cm9pbWFnZTwvc2Vjb25k

YXJ5LXRpdGxlPjwvdGl0bGVzPjxwZXJpb2RpY2FsPjxmdWxsLXRpdGxlPk5ldXJvaW1hZ2U8L2Z1

bGwtdGl0bGU+PGFiYnItMT5OZXVyb2ltYWdlPC9hYmJyLTE+PGFiYnItMj5OZXVyb2ltYWdlPC9h

YmJyLTI+PC9wZXJpb2RpY2FsPjxwYWdlcz4zMjAtNDE8L3BhZ2VzPjx2b2x1bWU+ODQ8L3ZvbHVt

ZT48a2V5d29yZHM+PGtleXdvcmQ+QWRvbGVzY2VudDwva2V5d29yZD48a2V5d29yZD5BZHVsdDwv

a2V5d29yZD48a2V5d29yZD5BbGdvcml0aG1zPC9rZXl3b3JkPjxrZXl3b3JkPipBcnRpZmFjdHM8

L2tleXdvcmQ+PGtleXdvcmQ+QnJhaW4vKnBoeXNpb2xvZ3k8L2tleXdvcmQ+PGtleXdvcmQ+QnJh

aW4gTWFwcGluZy8qbWV0aG9kczwva2V5d29yZD48a2V5d29yZD5GZW1hbGU8L2tleXdvcmQ+PGtl

eXdvcmQ+SGVhZCBNb3ZlbWVudHMvKnBoeXNpb2xvZ3k8L2tleXdvcmQ+PGtleXdvcmQ+SHVtYW5z

PC9rZXl3b3JkPjxrZXl3b3JkPkltYWdlIEVuaGFuY2VtZW50LyptZXRob2RzPC9rZXl3b3JkPjxr

ZXl3b3JkPkltYWdlIEludGVycHJldGF0aW9uLCBDb21wdXRlci1Bc3Npc3RlZC8qbWV0aG9kczwv

a2V5d29yZD48a2V5d29yZD5NYWduZXRpYyBSZXNvbmFuY2UgSW1hZ2luZy8qbWV0aG9kczwva2V5

d29yZD48a2V5d29yZD5NYWxlPC9rZXl3b3JkPjxrZXl3b3JkPk1vdGlvbjwva2V5d29yZD48a2V5

d29yZD5QYXR0ZXJuIFJlY29nbml0aW9uLCBBdXRvbWF0ZWQvbWV0aG9kczwva2V5d29yZD48a2V5

d29yZD5SZXByb2R1Y2liaWxpdHkgb2YgUmVzdWx0czwva2V5d29yZD48a2V5d29yZD5SZXN0L3Bo

eXNpb2xvZ3k8L2tleXdvcmQ+PGtleXdvcmQ+U2Vuc2l0aXZpdHkgYW5kIFNwZWNpZmljaXR5PC9r

ZXl3b3JkPjxrZXl3b3JkPlN1YnRyYWN0aW9uIFRlY2huaXF1ZTwva2V5d29yZD48a2V5d29yZD5Z

b3VuZyBBZHVsdDwva2V5d29yZD48a2V5d29yZD5BcnRpZmFjdDwva2V5d29yZD48a2V5d29yZD5G

dW5jdGlvbmFsIGNvbm5lY3Rpdml0eTwva2V5d29yZD48a2V5d29yZD5Ncmk8L2tleXdvcmQ+PGtl

eXdvcmQ+TW92ZW1lbnQ8L2tleXdvcmQ+PGtleXdvcmQ+UmVzdGluZyBzdGF0ZTwva2V5d29yZD48

L2tleXdvcmRzPjxkYXRlcz48eWVhcj4yMDE0PC95ZWFyPjxwdWItZGF0ZXM+PGRhdGU+SmFuIDE8

L2RhdGU+PC9wdWItZGF0ZXM+PC9kYXRlcz48aXNibj4xMDk1LTk1NzIgKEVsZWN0cm9uaWMpJiN4

RDsxMDUzLTgxMTkgKExpbmtpbmcpPC9pc2JuPjxhY2Nlc3Npb24tbnVtPjIzOTk0MzE0PC9hY2Nl

c3Npb24tbnVtPjx1cmxzPjxyZWxhdGVkLXVybHM+PHVybD5odHRwOi8vd3d3Lm5jYmkubmxtLm5p

aC5nb3YvcHVibWVkLzIzOTk0MzE0PC91cmw+PC9yZWxhdGVkLXVybHM+PC91cmxzPjxjdXN0b20y

PlBNQzM4NDkzMzg8L2N1c3RvbTI+PGVsZWN0cm9uaWMtcmVzb3VyY2UtbnVtPjEwLjEwMTYvai5u

ZXVyb2ltYWdlLjIwMTMuMDguMDQ4PC9lbGVjdHJvbmljLXJlc291cmNlLW51bT48L3JlY29yZD48

L0NpdGU+PC9FbmROb3RlPn==

ADDIN EN.CITE.DATA (Power et al., 2014).Machine learning models. In the present study, we used three machine learning methods - logistic regression (LR), support vector machine (SVM) and deep learning (DL) to perform single-subject classification. These methods were chosen in light of their widespread use amongst the neuroimaging community and their varying degrees of complexity and abstraction. In particular, LR is considered to be the “simplest” ML method; SVM is associated with medium levels of complexity and abstraction, and is the most widely used ML method in psychiatric neuroimaging; and DL is thought to have the greatest potential of detecting hidden patterns in the data, due to its higher levels of complexity and abstraction. These methods are described in detail below. Treatment of whole brain images. For LR and DL, the whole brain images were treated as 3D * time data. Here we treated each time point separately and computed the subject's final prediction based on the average probability across all time points; for SVM, the whole brain images were treated as 4D data. The different treatment of whole brain images was due to computational constrains (i.e. memory limitation and computing time) affecting the implementation of LR and DL but not SVM.Logistic Regression. The LR model fits a separating hyperplane, i.e. a linear function of the input features, between different classes (e.g. patients and controls). Within this framework, the probability that an observation belongs to a specific class is determined by the sigmoid function of a weighted sum of the input. These weights are estimated using a training sample and then assessed using an independent testing sample. In the present study, logistic regression was implemented using the Lasagne library (). We initialized the weights of the classifier using the Glorot method ADDIN EN.CITE <EndNote><Cite><Author>Glorot</Author><Year>2010</Year><RecNum>117</RecNum><DisplayText>(Glorot and Bengio, 2010)</DisplayText><record><rec-number>117</rec-number><foreign-keys><key app="EN" db-id="txazxa098zz5foe05vrxvsfhp5eeep9svrpt">117</key></foreign-keys><ref-type name="Journal Article">17</ref-type><contributors><authors><author>Glorot, Xavier</author><author>Bengio, Yoshua</author></authors></contributors><titles><title>Understanding the difficulty of training deep feedforward neural networks</title><secondary-title>Journal of Machine Learning Research</secondary-title></titles><periodical><full-title>Journal of Machine Learning Research</full-title></periodical><pages>249-256</pages><volume>9</volume><dates><year>2010</year></dates><urls></urls></record></Cite></EndNote>(Glorot and Bengio, 2010), with parameters sampled from the uniform distribution, and initialized the bias with zero value. These parameters were then optimized using the Rmsprop algorithm ADDIN EN.CITE <EndNote><Cite><Author>Tieleman</Author><Year>2012</Year><RecNum>2005</RecNum><DisplayText>(Tieleman and Hinton, 2012)</DisplayText><record><rec-number>2005</rec-number><foreign-keys><key app="EN" db-id="few0ewrtoapwveerppypr29sf20p2rdezdp5" timestamp="1508095503">2005</key></foreign-keys><ref-type name="Journal Article">17</ref-type><contributors><authors><author>Tieleman, Tijmen</author><author>Hinton, Geoffrey</author></authors></contributors><titles><title>Lecture 6.5-rmsprop: Divide the gradient by a running average of its recent magnitude</title><secondary-title>COURSERA: Neural networks for machine learning</secondary-title></titles><periodical><full-title>COURSERA: Neural networks for machine learning</full-title></periodical><pages>26-31</pages><volume>4</volume><number>2</number><dates><year>2012</year></dates><urls></urls></record></Cite></EndNote>(Tieleman and Hinton, 2012) using a mini-batch with 16 training samples, 100 training epochs, and a learning rate starting with a value of 0.05 with a learning rate decay decrease a factor of 10-6 per epoch. An L2 penalization with parameter value equal to 5*105 was applied to the model to penalize weights with excessively high values and to prevent overfitting.Support Vector Machine. The SVM model ADDIN EN.CITE <EndNote><Cite><Author>Cortes</Author><Year>1995</Year><RecNum>7</RecNum><DisplayText>(Cortes and Vapnik, 1995)</DisplayText><record><rec-number>7</rec-number><foreign-keys><key app="EN" db-id="0traede992tpe8erpx8pxxz3s202sdfaazva" timestamp="1547146135">7</key></foreign-keys><ref-type name="Journal Article">17</ref-type><contributors><authors><author>Cortes, C.</author><author>Vapnik, V.</author></authors></contributors><titles><title>Support Vector Network</title><secondary-title>Machine Learning</secondary-title></titles><pages><style face="normal" font="default" size="100%">273</style><style face="normal" font="default" charset="134" size="100%">-297</style></pages><volume>20</volume><number>3</number><dates><year>1995</year></dates><urls></urls></record></Cite></EndNote>(Cortes and Vapnik, 1995) maps the input vectors to a feature space using a set of mathematical functions known as kernels. In this feature space, the model finds the optimum separation surface that can maximise the margin between different classes within a training dataset. Once the separation surface is determined, it can be used to predict the class of new observations using an independent testing dataset. Here a linear kernel was preferred to a nonlinear one to minimise the risk of overfitting. The model was based on libsvm and implemented using the Scikit-Learn library ADDIN EN.CITE <EndNote><Cite><Author>Pedregosa</Author><Year>2012</Year><RecNum>116</RecNum><DisplayText>(Pedregosa<style face="italic"> et al.</style>, 2012)</DisplayText><record><rec-number>116</rec-number><foreign-keys><key app="EN" db-id="txazxa098zz5foe05vrxvsfhp5eeep9svrpt">116</key></foreign-keys><ref-type name="Journal Article">17</ref-type><contributors><authors><author>Pedregosa, Fabian</author><author>Varoquaux, Ga</author><author>Gramfort, Alexandre</author><author>Michel, Vincent</author><author>Thirion, Bertrand</author><author>Grisel, Olivier</author><author>Blondel, Mathieu</author><author>Prettenhofer, Peter</author><author>Weiss, Ron</author><author>Dubourg, Vincent</author></authors></contributors><titles><title>Scikit-learn: Machine Learning in Python</title><secondary-title>Journal of Machine Learning Research</secondary-title></titles><periodical><full-title>Journal of Machine Learning Research</full-title></periodical><pages>2825-2830</pages><volume>12</volume><number>10</number><dates><year>2012</year></dates><urls></urls></record></Cite></EndNote>(Pedregosa et al., 2012). The linear SVM has only one parameter (soft margin parameter C) that controls the trade-off between reducing training errors and having a larger separation margin. This parameter was optimized spitting the data into training and validation sets and performing a grid search (i.e., C = 2-5, 2-3, 2-1, 2-0, 21, 23 … 213, 215) to estimate the best value.Deep Learning Technology. DL models are a set of machine learning algorithms that extract multiple levels features from the input data PEVuZE5vdGU+PENpdGU+PEF1dGhvcj5WaWVpcmE8L0F1dGhvcj48WWVhcj4yMDE3PC9ZZWFyPjxS

ZWNOdW0+MTQ2PC9SZWNOdW0+PERpc3BsYXlUZXh0PihWaWVpcmE8c3R5bGUgZmFjZT0iaXRhbGlj

Ij4gZXQgYWwuPC9zdHlsZT4sIDIwMTcpPC9EaXNwbGF5VGV4dD48cmVjb3JkPjxyZWMtbnVtYmVy

PjE0NjwvcmVjLW51bWJlcj48Zm9yZWlnbi1rZXlzPjxrZXkgYXBwPSJFTiIgZGItaWQ9InJkNTV6

MmV2MGVmZHJtZTUyZmFwMnp2NDB2cHp0Zjk5OWFkZiIgdGltZXN0YW1wPSIxNTQ3MTQ1Mzc1Ij4x

NDY8L2tleT48L2ZvcmVpZ24ta2V5cz48cmVmLXR5cGUgbmFtZT0iSm91cm5hbCBBcnRpY2xlIj4x

NzwvcmVmLXR5cGU+PGNvbnRyaWJ1dG9ycz48YXV0aG9ycz48YXV0aG9yPlZpZWlyYSwgUy48L2F1

dGhvcj48YXV0aG9yPlBpbmF5YSwgVy4gSC48L2F1dGhvcj48YXV0aG9yPk1lY2hlbGxpLCBBLjwv

YXV0aG9yPjwvYXV0aG9ycz48L2NvbnRyaWJ1dG9ycz48YXV0aC1hZGRyZXNzPkRlcGFydG1lbnQg

b2YgUHN5Y2hvc2lzIFN0dWRpZXMsIEluc3RpdHV0ZSBvZiBQc3ljaGlhdHJ5LCBQc3ljaG9sb2d5

ICZhbXA7IE5ldXJvc2NpZW5jZSwgS2luZyZhcG9zO3MgQ29sbGVnZSBMb25kb24sIDE2IERlIENy

ZXNwaWdueSBQYXJrLCBTRTUgOEFGLCBVbml0ZWQgS2luZ2RvbS4gRWxlY3Ryb25pYyBhZGRyZXNz

OiBzYW5kcmEudmllaXJhQGtjbC5hYy51ay4mI3hEO0NlbnRyZSBvZiBNYXRoZW1hdGljcywgQ29t

cHV0YXRpb24sIGFuZCBDb2duaXRpb24sIFVuaXZlcnNpZGFkZSBGZWRlcmFsIGRvIEFCQywgUnVh

IEFyY3R1cnVzLCBKYXJkaW0gQW50YXJlcywgU2FvIEJlcm5hcmRvIGRvIENhbXBvLCBTUCBDRVAg

MDkuNjA2LTA3MCwgQnJhemlsLiYjeEQ7RGVwYXJ0bWVudCBvZiBQc3ljaG9zaXMgU3R1ZGllcywg

SW5zdGl0dXRlIG9mIFBzeWNoaWF0cnksIFBzeWNob2xvZ3kgJmFtcDsgTmV1cm9zY2llbmNlLCBL

aW5nJmFwb3M7cyBDb2xsZWdlIExvbmRvbiwgMTYgRGUgQ3Jlc3BpZ255IFBhcmssIFNFNSA4QUYs

IFVuaXRlZCBLaW5nZG9tLjwvYXV0aC1hZGRyZXNzPjx0aXRsZXM+PHRpdGxlPlVzaW5nIGRlZXAg

bGVhcm5pbmcgdG8gaW52ZXN0aWdhdGUgdGhlIG5ldXJvaW1hZ2luZyBjb3JyZWxhdGVzIG9mIHBz

eWNoaWF0cmljIGFuZCBuZXVyb2xvZ2ljYWwgZGlzb3JkZXJzOiBNZXRob2RzIGFuZCBhcHBsaWNh

dGlvbnM8L3RpdGxlPjxzZWNvbmRhcnktdGl0bGU+TmV1cm9zY2kgQmlvYmVoYXYgUmV2PC9zZWNv

bmRhcnktdGl0bGU+PC90aXRsZXM+PHBlcmlvZGljYWw+PGZ1bGwtdGl0bGU+TmV1cm9zY2llbmNl

IGFuZCBCaW9iZWhhdmlvcmFsIFJldmlld3M8L2Z1bGwtdGl0bGU+PGFiYnItMT5OZXVyb3NjaS4g

QmlvYmVoYXYuIFJldi48L2FiYnItMT48YWJici0yPk5ldXJvc2NpIEJpb2JlaGF2IFJldjwvYWJi

ci0yPjxhYmJyLTM+TmV1cm9zY2llbmNlICZhbXA7IEJpb2JlaGF2aW9yYWwgUmV2aWV3czwvYWJi

ci0zPjwvcGVyaW9kaWNhbD48cGFnZXM+NTgtNzU8L3BhZ2VzPjx2b2x1bWU+NzQ8L3ZvbHVtZT48

bnVtYmVyPlB0IEE8L251bWJlcj48a2V5d29yZHM+PGtleXdvcmQ+SHVtYW5zPC9rZXl3b3JkPjxr

ZXl3b3JkPk1hY2hpbmUgTGVhcm5pbmc8L2tleXdvcmQ+PGtleXdvcmQ+Kk1lbnRhbCBEaXNvcmRl

cnM8L2tleXdvcmQ+PGtleXdvcmQ+Kk5lcnZvdXMgU3lzdGVtIERpc2Vhc2VzPC9rZXl3b3JkPjxr

ZXl3b3JkPk5ldXJhbCBOZXR3b3JrcyAoQ29tcHV0ZXIpPC9rZXl3b3JkPjxrZXl3b3JkPk5ldXJv

aW1hZ2luZzwva2V5d29yZD48a2V5d29yZD5TcGVlY2g8L2tleXdvcmQ+PGtleXdvcmQ+KkF1dG9l

bmNvZGVyczwva2V5d29yZD48a2V5d29yZD4qQ29udm9sdXRpb25hbCBuZXVyYWwgbmV0d29ya3M8

L2tleXdvcmQ+PGtleXdvcmQ+KkRlZXAgYmVsaWVmIG5ldHdvcmtzPC9rZXl3b3JkPjxrZXl3b3Jk

PipEZWVwIGxlYXJuaW5nPC9rZXl3b3JkPjxrZXl3b3JkPipNYWNoaW5lIGxlYXJuaW5nPC9rZXl3

b3JkPjxrZXl3b3JkPipNdWx0aWxheWVyIHBlcmNlcHRyb248L2tleXdvcmQ+PGtleXdvcmQ+Kk5l

dXJvaW1hZ2luZzwva2V5d29yZD48a2V5d29yZD4qTmV1cm9sb2dpYyBkaXNvcmRlcnM8L2tleXdv

cmQ+PGtleXdvcmQ+KlBhdHRlcm4gcmVjb2duaXRpb248L2tleXdvcmQ+PGtleXdvcmQ+KlBzeWNo

aWF0cmljIGRpc29yZGVyczwva2V5d29yZD48L2tleXdvcmRzPjxkYXRlcz48eWVhcj4yMDE3PC95

ZWFyPjxwdWItZGF0ZXM+PGRhdGU+TWFyPC9kYXRlPjwvcHViLWRhdGVzPjwvZGF0ZXM+PGlzYm4+

MTg3My03NTI4IChFbGVjdHJvbmljKSYjeEQ7MDE0OS03NjM0IChMaW5raW5nKTwvaXNibj48YWNj

ZXNzaW9uLW51bT4yODA4NzI0MzwvYWNjZXNzaW9uLW51bT48dXJscz48cmVsYXRlZC11cmxzPjx1

cmw+aHR0cDovL3d3dy5uY2JpLm5sbS5uaWguZ292L3B1Ym1lZC8yODA4NzI0MzwvdXJsPjwvcmVs

YXRlZC11cmxzPjwvdXJscz48ZWxlY3Ryb25pYy1yZXNvdXJjZS1udW0+MTAuMTAxNi9qLm5ldWJp

b3Jldi4yMDE3LjAxLjAwMjwvZWxlY3Ryb25pYy1yZXNvdXJjZS1udW0+PC9yZWNvcmQ+PC9DaXRl

PjwvRW5kTm90ZT5=

ADDIN EN.CITE PEVuZE5vdGU+PENpdGU+PEF1dGhvcj5WaWVpcmE8L0F1dGhvcj48WWVhcj4yMDE3PC9ZZWFyPjxS

ZWNOdW0+MTQ2PC9SZWNOdW0+PERpc3BsYXlUZXh0PihWaWVpcmE8c3R5bGUgZmFjZT0iaXRhbGlj

Ij4gZXQgYWwuPC9zdHlsZT4sIDIwMTcpPC9EaXNwbGF5VGV4dD48cmVjb3JkPjxyZWMtbnVtYmVy

PjE0NjwvcmVjLW51bWJlcj48Zm9yZWlnbi1rZXlzPjxrZXkgYXBwPSJFTiIgZGItaWQ9InJkNTV6

MmV2MGVmZHJtZTUyZmFwMnp2NDB2cHp0Zjk5OWFkZiIgdGltZXN0YW1wPSIxNTQ3MTQ1Mzc1Ij4x

NDY8L2tleT48L2ZvcmVpZ24ta2V5cz48cmVmLXR5cGUgbmFtZT0iSm91cm5hbCBBcnRpY2xlIj4x

NzwvcmVmLXR5cGU+PGNvbnRyaWJ1dG9ycz48YXV0aG9ycz48YXV0aG9yPlZpZWlyYSwgUy48L2F1

dGhvcj48YXV0aG9yPlBpbmF5YSwgVy4gSC48L2F1dGhvcj48YXV0aG9yPk1lY2hlbGxpLCBBLjwv

YXV0aG9yPjwvYXV0aG9ycz48L2NvbnRyaWJ1dG9ycz48YXV0aC1hZGRyZXNzPkRlcGFydG1lbnQg

b2YgUHN5Y2hvc2lzIFN0dWRpZXMsIEluc3RpdHV0ZSBvZiBQc3ljaGlhdHJ5LCBQc3ljaG9sb2d5

ICZhbXA7IE5ldXJvc2NpZW5jZSwgS2luZyZhcG9zO3MgQ29sbGVnZSBMb25kb24sIDE2IERlIENy

ZXNwaWdueSBQYXJrLCBTRTUgOEFGLCBVbml0ZWQgS2luZ2RvbS4gRWxlY3Ryb25pYyBhZGRyZXNz

OiBzYW5kcmEudmllaXJhQGtjbC5hYy51ay4mI3hEO0NlbnRyZSBvZiBNYXRoZW1hdGljcywgQ29t

cHV0YXRpb24sIGFuZCBDb2duaXRpb24sIFVuaXZlcnNpZGFkZSBGZWRlcmFsIGRvIEFCQywgUnVh

IEFyY3R1cnVzLCBKYXJkaW0gQW50YXJlcywgU2FvIEJlcm5hcmRvIGRvIENhbXBvLCBTUCBDRVAg

MDkuNjA2LTA3MCwgQnJhemlsLiYjeEQ7RGVwYXJ0bWVudCBvZiBQc3ljaG9zaXMgU3R1ZGllcywg

SW5zdGl0dXRlIG9mIFBzeWNoaWF0cnksIFBzeWNob2xvZ3kgJmFtcDsgTmV1cm9zY2llbmNlLCBL

aW5nJmFwb3M7cyBDb2xsZWdlIExvbmRvbiwgMTYgRGUgQ3Jlc3BpZ255IFBhcmssIFNFNSA4QUYs

IFVuaXRlZCBLaW5nZG9tLjwvYXV0aC1hZGRyZXNzPjx0aXRsZXM+PHRpdGxlPlVzaW5nIGRlZXAg

bGVhcm5pbmcgdG8gaW52ZXN0aWdhdGUgdGhlIG5ldXJvaW1hZ2luZyBjb3JyZWxhdGVzIG9mIHBz

eWNoaWF0cmljIGFuZCBuZXVyb2xvZ2ljYWwgZGlzb3JkZXJzOiBNZXRob2RzIGFuZCBhcHBsaWNh

dGlvbnM8L3RpdGxlPjxzZWNvbmRhcnktdGl0bGU+TmV1cm9zY2kgQmlvYmVoYXYgUmV2PC9zZWNv

bmRhcnktdGl0bGU+PC90aXRsZXM+PHBlcmlvZGljYWw+PGZ1bGwtdGl0bGU+TmV1cm9zY2llbmNl

IGFuZCBCaW9iZWhhdmlvcmFsIFJldmlld3M8L2Z1bGwtdGl0bGU+PGFiYnItMT5OZXVyb3NjaS4g

QmlvYmVoYXYuIFJldi48L2FiYnItMT48YWJici0yPk5ldXJvc2NpIEJpb2JlaGF2IFJldjwvYWJi

ci0yPjxhYmJyLTM+TmV1cm9zY2llbmNlICZhbXA7IEJpb2JlaGF2aW9yYWwgUmV2aWV3czwvYWJi

ci0zPjwvcGVyaW9kaWNhbD48cGFnZXM+NTgtNzU8L3BhZ2VzPjx2b2x1bWU+NzQ8L3ZvbHVtZT48

bnVtYmVyPlB0IEE8L251bWJlcj48a2V5d29yZHM+PGtleXdvcmQ+SHVtYW5zPC9rZXl3b3JkPjxr

ZXl3b3JkPk1hY2hpbmUgTGVhcm5pbmc8L2tleXdvcmQ+PGtleXdvcmQ+Kk1lbnRhbCBEaXNvcmRl

cnM8L2tleXdvcmQ+PGtleXdvcmQ+Kk5lcnZvdXMgU3lzdGVtIERpc2Vhc2VzPC9rZXl3b3JkPjxr

ZXl3b3JkPk5ldXJhbCBOZXR3b3JrcyAoQ29tcHV0ZXIpPC9rZXl3b3JkPjxrZXl3b3JkPk5ldXJv

aW1hZ2luZzwva2V5d29yZD48a2V5d29yZD5TcGVlY2g8L2tleXdvcmQ+PGtleXdvcmQ+KkF1dG9l

bmNvZGVyczwva2V5d29yZD48a2V5d29yZD4qQ29udm9sdXRpb25hbCBuZXVyYWwgbmV0d29ya3M8

L2tleXdvcmQ+PGtleXdvcmQ+KkRlZXAgYmVsaWVmIG5ldHdvcmtzPC9rZXl3b3JkPjxrZXl3b3Jk

PipEZWVwIGxlYXJuaW5nPC9rZXl3b3JkPjxrZXl3b3JkPipNYWNoaW5lIGxlYXJuaW5nPC9rZXl3

b3JkPjxrZXl3b3JkPipNdWx0aWxheWVyIHBlcmNlcHRyb248L2tleXdvcmQ+PGtleXdvcmQ+Kk5l

dXJvaW1hZ2luZzwva2V5d29yZD48a2V5d29yZD4qTmV1cm9sb2dpYyBkaXNvcmRlcnM8L2tleXdv

cmQ+PGtleXdvcmQ+KlBhdHRlcm4gcmVjb2duaXRpb248L2tleXdvcmQ+PGtleXdvcmQ+KlBzeWNo

aWF0cmljIGRpc29yZGVyczwva2V5d29yZD48L2tleXdvcmRzPjxkYXRlcz48eWVhcj4yMDE3PC95

ZWFyPjxwdWItZGF0ZXM+PGRhdGU+TWFyPC9kYXRlPjwvcHViLWRhdGVzPjwvZGF0ZXM+PGlzYm4+

MTg3My03NTI4IChFbGVjdHJvbmljKSYjeEQ7MDE0OS03NjM0IChMaW5raW5nKTwvaXNibj48YWNj

ZXNzaW9uLW51bT4yODA4NzI0MzwvYWNjZXNzaW9uLW51bT48dXJscz48cmVsYXRlZC11cmxzPjx1

cmw+aHR0cDovL3d3dy5uY2JpLm5sbS5uaWguZ292L3B1Ym1lZC8yODA4NzI0MzwvdXJsPjwvcmVs

YXRlZC11cmxzPjwvdXJscz48ZWxlY3Ryb25pYy1yZXNvdXJjZS1udW0+MTAuMTAxNi9qLm5ldWJp

b3Jldi4yMDE3LjAxLjAwMjwvZWxlY3Ryb25pYy1yZXNvdXJjZS1udW0+PC9yZWNvcmQ+PC9DaXRl

PjwvRW5kTm90ZT5=

ADDIN EN.CITE.DATA (Vieira et al., 2017). Here higher-level features are learned as a non-linear combination of lower-level features, allowing the extraction of complex and abstract patterns in the data. Based on these higher-level features, the model determines the separation surface to perform the classification task. Here we used two types of deep learning models: the Convolutional Neural Network (ConvNet) and a deep neural network with fully connected layers.Because whole brain images data are associated with high dimensionality (over 50,000 voxels), the application of a deep neural network with fully connected layers to this type of data would require excessive computational resources. However, since whole brain time series data are also associated with correlation between neighbouring voxels, we used ConvNet (where the units are not connected to all the units in the previous layer) to reduce the number of parameters in the network. This reduction had the impact of reducing risk of overfitting whilst alleviating computational demands. The training of the ConvNet was performed using 3D kernels in which each time step was considered an input channel; training was implemented using the Rmsprop algorithm with a mini-batch with 16 training samples, 100 training epochs and a learning rate starting with a value of 0.05 and decaying a factor of 10-6 per epoch. The batch normalization ADDIN EN.CITE <EndNote><Cite><Author>Ioffe</Author><Year>2015</Year><RecNum>118</RecNum><DisplayText>(Ioffe and Szegedy, 2015)</DisplayText><record><rec-number>118</rec-number><foreign-keys><key app="EN" db-id="txazxa098zz5foe05vrxvsfhp5eeep9svrpt">118</key></foreign-keys><ref-type name="Journal Article">17</ref-type><contributors><authors><author>Ioffe, Sergey</author><author>Szegedy, Christian</author></authors></contributors><titles><title>Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift</title><secondary-title>Computer Science</secondary-title></titles><periodical><full-title>Computer Science</full-title></periodical><dates><year>2015</year></dates><urls></urls></record></Cite></EndNote>(Ioffe and Szegedy, 2015) was applied to all layers to accelerate the convergence of the training. The architecture of the deep learning model is represented in Figure S1.In contrast, it was possible to use a deep neural network with fully connected layers to analyse connectome-wide matrices and graph-based analytic metrics, in light of the lower dimensionality of these measures. In particular, we adopted a structure with three hidden layers with 100 units in each hidden layer. We used the Rectified Linear Unit (ReLU) in all hidden layers and the softmax units in the output layer. The training of the network was performed using the Rmsprop algorithm using a mini-batch with 16 training samples, 100 training epochs, and a learning rate starting with a value of 0.05 with a learning rate decay decrease a factor of 10-6 per epoch. The L2 parameter was set to 5*10-5. Both deep learning models were implemented using the Lasagne library.References ADDIN EN.REFLIST Andreasen, N. C., Flaum, M. & Arndt, S. (1992). The Comprehensive Assessment of Symptoms and History (CASH). An instrument for assessing diagnosis and psychopathology. Archives of General Psychiatry 49, 615-23.Association, A. P. (2000). Diagnostic and Statistical Manual of Mental Disorders fourth edition, Text revision (DSM-IV TR).Cortes, C. & Vapnik, V. (1995). Support Vector Network. Machine Learning 20, 273-297.First, M. B., Frances, A. & Pincus, H. A. (2004). DSM-IV-TR guidebook. American Psychiatric Pub.First, M. B., Spitzer, R. L., Gibbon, M., Williams, J. B. W., Spitzer, R., Gibbons, M. & Williams, J. (2012). Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I/P). John Wiley & Sons, Inc.Glorot, X. & Bengio, Y. (2010). Understanding the difficulty of training deep feedforward neural networks. Journal of Machine Learning Research 9, 249-256.Ioffe, S. & Szegedy, C. (2015). Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift. Computer Science.Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R. & Dubourg, V. (2012). Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research 12, 2825-2830.Poldrack, R. A., Congdon, E., Triplett, W., Gorgolewski, K. J., Karlsgodt, K. H., Mumford, J. A., Sabb, F. W., Freimer, N. B., London, E. D., Cannon, T. D. & Bilder, R. M. (2016). A phenome-wide examination of neural and cognitive function. Scientific Data 3, 160110.Power, J. D., Mitra, A., Laumann, T. O., Snyder, A. Z., Schlaggar, B. L. & Petersen, S. E. (2014). Methods to detect, characterize, and remove motion artifact in resting state fMRI. Neuroimage 84, 320-41.Tieleman, T. & Hinton, G. (2012). Lecture 6.5-rmsprop: Divide the gradient by a running average of its recent magnitude. COURSERA: Neural networks for machine learning 4, 26-31.Vieira, S., Pinaya, W. H. & Mechelli, A. (2017). Using deep learning to investigate the neuroimaging correlates of psychiatric and neurological disorders: Methods and applications. Neuroscience and Biobehavioral Reviews 74, 58-75.Yan, C. G., Craddock, R. C., He, Y. & Milham, M. P. (2013). Addressing head motion dependencies for small-world topologies in functional connectomics. Frontiers in Human Neuroscience 7, 910.Figure legendsFigure S1. ?The architecture of the deep learning model. We used a convolutional neural network in which the first layer was a 3D convolutional layer with 64 filters (each filter with size 7x7x7 voxels), a stride with size of 2, and "valid" padding (V). Next, a max pooling layer was used to reduce the dimensionality of the feature maps of the network. This layer had filter sizes of 5x5x5 voxels and stride size of 3. The third layer was a convolutional layer with 64 filters (each filter with size 5x5x5 voxels), a stride size 1, and a "same" padding. Next, we used a max pooling layer with filter size of 5x5x5 voxels and a stride size of 3 voxels. Finally, the feature maps were flattened from a 3D space to a 1D vector in order to be used as input to the next layer: a full-connected (FC) layer. This layer had 50 rectified linear units (ReLU). Lastly, the output layer had 2 units with softmax activation function that outputs the probability of the input data being from a patient or control. ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download