. 2021 Jun 16;11:12686. doi: 10.1038/s41598-021-92155-6

Table 3.

Summary of the feature selection and modeling approaches for each analysis (DAFIT Filt analyses were only fit to the combined masks).

	Number of potential features		Lincomb used	Corr thresholds Single mask /combined mask	PCA thresholds Single mask/combined mask
	LV mask	RV mask	Lincomb used	Corr thresholds Single mask /combined mask	PCA thresholds Single mask/combined mask
Original	348	348	Yes	0.6/0.5	0.9/0.85
ICC2	22	46	No	0.9/0.9	0.9/0.9
ICC3	8	24	No	–	–
DAFIT	348	348	Yes	0.6/0.4	0.9/0.85
DAFIT Filt2	22	46	No	–/0.8	–/0.85
DAFIT Filt3	8	24	No	–/0.8	–/0.85

Original original data without inclusion of side experiments, ICC2 features with excellent intraclass correlation from first two extractions, ICC3 features with excellent intraclass correlation from all three extractions, DAFIT synthetic data creation using main and side study data, DAFIT Filt2 combining DAFIT with feature filtering from ICC2, DAFIT Filt3 combining DAFIT with feature filtering from ICC3, full full feature set, corr high correlation filter, pca principal component analysis, lincomb linear combinations filter, LV left ventricle, RV right ventricle.