Table 3.

Results for all classification experiments.

Method		Macro average			Micro average
		Precision	Recall	F1-score	Precision	Recall	F1-score	MCC
Normal CEL	DensNet-161 (fold 0)	0.2165	0.2341	0.1923	0.7375	0.7375	0.7375	0.3707
	DensNet-161 (fold 1)	0.3493	0.3158	0.2996	0.7327	0.7327	0.7327	0.4604
	Avereage	0.2829	0.2749	0.2459	0.7351	0.7351	0.7351	0.4156
	ResNet-152 (fold 0)	0.3302	0.2401	0.1970	0.7203	0.7203	0.7203	0.3520
	ResNet-152 (fold 1)	0.3431	0.2805	0.2789	0.7481	0.7481	0.7481	0.4718
	Average	0.3367	0.2603	0.2379	0.7342	0.7342	0.7342	0.4119
Weighted CEL	DensNet-161 (fold 0)	0.2933	0.2939	0.2523	0.7195	0.7195	0.7195	0.3998
	DensNet-161 (fold 1)	0.3163	0.2914	0.2581	0.6991	0.6991	0.6991	0.4054
	Average	0.3048	0.2927	0.2552	0.7093	0.7093	0.7093	0.4026
	ResNet-152 (fold 0)	0.2136	0.2872	0.2186	0.6568	0.6568	0.6568	0.3588
	ResNet-152 (fold 1)	0.3033	0.2799	0.2478	0.6890	0.6890	0.6890	0.3966
	Average	0.2585	0.2836	0.2332	0.6729	0.6729	0.6729	0.3777
Weighted sampling	DensNet-161 (fold 0)	0.2525	0.2794	0.2315	0.7332	0.7332	0.7332	0.4111
	DensNet-161 (fold 1)	0.3463	0.2830	0.2806	0.7400	0.7400	0.7400	0.4547
	Average	0.2994	0.2812	0.2560	0.7366	0.7366	0.7366	0.4329
	ResNet-152 (fold 0)	0.2637	0.2930	0.2334	0.7324	0.7324	0.7324	0.4088
	ResNet-152 (fold 1)	0.3088	0.2619	0.2417	0.7316	0.7316	0.7316	0.4520
	Average	0.2862	0.2774	0.2375	0.7320	0.7320	0.7320	0.4304

Experiments were done with and without weighted cross-entropy loss (CEL) and using a weighted sampling technique. Bold numbers represent the best average value of that column.