Table 3.

The sensitivities and FROC scores for faster R-CNN and YOLO v2 detections of C4d positive and negative PTC with different detection models trained by different dataset at different mean number of false positives per feasible ROIs (0 to 2 and 0 to 8 for detection of positive and negative PTC, respectively). Model 1: trained by subset 1, Model 2: trained by subset 2, Model 3: trained by fusion of subset 1 and 2.

Mean of FPs	Detection model for C4d positive PTC			Detection model for C4d negative PTC
	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3
Faster R-CNN
0.125	0.6970	0.5495	0.5768	0.1387	0.0863	0.1148
0.250	0.7803	0.6923	0.7510	0.3333	0.1969	0.2405
0.500	0.7886	0.8791	0.8817	0.3966	0.4579	0.4343
1.000	0.9024	0.9451	0.9253	0.5615	0.6969	0.6644
2.000	0.9187	0.9478	0.9585	0.7082	0.8424	0.8131
4.000	0.9187	0.9478	0.9647	0.8075	0.9294	0.8887
8.000	0.9187	0.9478	0.9647	0.8563	0.9294	0.8910
Score	0.8463	0.8442	0.8603	0.5431	0.5913	0.5781
YOLO v2
0.125	0.3864	0.7009	0.6736	0.0058	0.2034	0.1928
0.250	0.6284	0.7479	0.6795	0.0032	0.3644	0.2240
0.500	0.6817	0.8333	0.7329	0.2945	0.5512	0.4494
1.000	0.7124	0.8462	0.7567	0.5394	0.6617	0.4761
2.000	0.7221	0.8547	0.7565	0.5423	0.7480	0.6121
4.000	0.7444	0. 8761	0.7864	0.5423	0.8100	0.7106
8.000	0.7444	0.8846	0.7864	0.5423	0.8100	0.7345
Score	0.6599	0.8112	0.7388	0.3528	0.5926	0.4856