Table 3. Performances of the SHG/TPEF fibrosis scoring models (AUCs) (generalized linear model vs. SVM) using different combinations of morphological features.

ISHAK fibrosis stages	Features used for modeling
	F1-3		F1-13		F1-19
	AUC	P	AUC	P	AUC	P
Generalized Linear Model
1 vs. 23456	0.595	0.34	0.597	0.33	0.640	0.16
12 vs. 3456	0.552	0.49	0.608	0.15	0.628	0.09
123 vs. 456	0.697	0.01	0.687	0.01	0.788	<0.01
1234 vs. 56	0.745	<0.01	0.788	<0.01	0.829	<0.01
12345 vs. 6	0.736	0.17	0.408	0.59	0.826	0.06
SVM (RBF kernel) model
1 vs. 23456	0.587	0.38	0.735	0.02	0.745	0.01
12 vs. 3456	0.548	0.52	0.641	0.06	0.774	<0.01
123 vs. 456	0.697	0.01	0.813	<0.01	0.797	<0.01
1234 vs. 56	0.745	<0.01	0.795	<0.01	0.814	<0.01
12345 vs. 6	0.736	0.17	0.726	0.19	0.806	0.07