Table 1.

List of single-site studies that provided machine learning model for predicting schizophrenia using resting-state brain patterns

Study	Year	Total: Size of classes	Accuracy
Shen et al.83	2010	52: 32 SCZ, 20 HC	86.50%
Fan et al.84	2011	62: 31 SCZ, 31 HC	87.1%a
Yu et al.85	2013	89: 32 SCZ, 38 HC (+19 MDD)	80.9%
Anderson and Cohen86	2013	146: 74 SCZ, 72 HC (COBRE dataset)	65%
Arbabshirani et al.87	2013	56: 28 SCZ, 28 HC	96%a
Yu et al.88	2013	71: 24 SCZ, 25 healthy siblings of SCZ, 22 HC	62%
Guo et al.89	2014	131: 69 SCZ, 62 HC	80%
Brodersen et al.90	2014	83: 41 SCZ, 42 HC	78%a
Anticevic et al.91	2014	180: 90 SCZ, 90 HC	73.9%
Watanabe et al.92	2014	123: 54 SCZ, 67 HC	73.50%
Chyzhyk et al.93	2015	54: 26 SCZ with history of AH, 14 SCZ without a history of AH, 28 HC	97.1%a
Cheng et al.94	2015	48: 19 SCZ, 29 HC	79%
Peters et al.95	2016	36: 18 SCZ, 18 HC	91%a
Mikolas et al.96	2016	126: 63 SCZ with FE SCZ, 63 HC	73%
Cabral et al.34	2016	132: 66 SCZ, 66 HC (COBRE dataset)	70.5%
Yang et al.97	2016	86: 40 SCZ, 46 HC	77.91%
Iwabuchi and Palaniyappan98	2017	133: 62 SCZ, 71 HC	78.04%a
Lottman et al.99	2017	69: 34 unmedicated (17 drug-naive) SCZ + follow-up post treatment, 35 HC	83.8%a
Guo et al.100	2017	68: 28 FE drug-naive SCZ, 28 family-based controls, 40 HC	92.86%a

SCZ Schizophrenia, HC Healthy controls, AH Auditory hallucinations, FE First episode, MDD Major depression

^aAccuracy of best model among several reported models