. Author manuscript; available in PMC: 2018 Sep 1.

Published in final edited form as: Mach Learn Med Imaging. 2017 Sep 7;10541:371–378. doi: 10.1007/978-3-319-67389-9_43

Table 2.

Test performance of the models with the best F-score_mod on evaluation (TRAIN EVAL). Excepting the thalamus, overall models’ performance generalizes to out-of-sample test data

ROI	Model	CV	Features	Thres	Eval F-recall	Eval F-share	Eval F-score	Test F-recall	Test F-share	Test F-score
Accumbens	GBDT	5-fold	Normed	0.014	0.83	0.2	0.81	0.75	0.27	0.74
Amygdala	GBDT	5-fold	Normed	0.007	0.89	0.4	0.72	0.76	0.40	0.67
Caudate	SVC	LOSO	Original	0.017	0.84	0.3	0.76	0.92	0.45	0.68
Hippocampus	GBDT	5-fold	Normed	0.010	0.85	0.2	0.83	0.91	0.29	0.80
Pallidum	GBDT	5-fold	Normed	0.009	0.86	0.2	0.83	0.86	0.30	0.77
Putamen	SVC	LOSO	Normed	0.007	0.84	0.4	0.70	0.65	0.44	0.60
Thalamus	SVC	LOSO	Original	0.007	0.84	0.4	0.70	1.00	1.00	0.00