Representation of Differential Learning Method for Mitosis Detection

Table 1

Mitosis detection performance with the proposed method on the Aperio-type images.

Method	α	γ	k	m	Precision	Recall	F1-score
FPN baseline	—	—	—	—	0.578	0.599	0.589
FPN-FL	0.5	0	—	—	0.586	0.613	0.595
FPN-FL	0.5	0.5	—	—	0.614	0.645	0.629
FPN-FL	0.5	1	—	—	0.563	0.584	0.573
FPN-FL	0.5	2	—	—	0.496	0.647	0.561
GLB-FPN	—	—	256	256	0.614	0.675	0.643
GLB-FPN	—	—	256	512	0.647	0.685	0.662
RDLM	0.5	0	256	256	0.623	0.654	0.638
RDLM	0.5	0.5	256	256	0.685	0.70	0.692
RDLM	0.5	1	256	256	0.591	0.640	0.614
RDLM	0.5	2	256	256	0.546	0.674	0.602