Table 1.

Publications on using supervised ML algorithms on fMRI data for distinguishing ASD from TD patients. Based on²⁰.

Authors	Data size	ML methods	AUC	Accuracy	Recall	Precision
21	$\text{n}=505$ ASD ASD; $\text{n}=530$ TD (ABIDE)	Deep learning	–	0.95	0.97	–
22	$\text{n}=462$ ASD; $\text{n}=464$ TD (ABIDE)	Deep learning	0.96	0.87	0.87	–
23	$\text{n}=126$ ASD; $\text{n}=126$ TD (ABIDE)	RF	0.91	0.89	0.93	–
24	(ABIDE)	SVM	–	0.80	–	–
25	$\text{n}=15$ ASD; $\text{n}=45$ TD (ABIDE)	Lasso	0.78	0.75	0.77	0.73
26	$\text{n}=505$ ASD; $\text{n}=530$ TD (ABIDE)	Deep learning	–	0.70	0.74	0.63
27	$\text{n}=505$ ASD; $\text{n}=530$ TD	Deep belief network	0.76	–	–	–
28	$\text{n}=59$ ASD; $\text{n}=59$ TD	Linear SVM; ridge logistic regression	0.73	–	–	–