TSHVNet: Simultaneous Nuclear Instance Segmentation and Classification in Histopathological Images Based on Multiattention Mechanisms

Table 4

Evaluation results of nuclear segmentation on different combinations of SimAM modules.

Method	Dataset metrics
	CoNSeP					PanNuke
	DICE	AJI	DQ	SQ	PQ	DICE	AJI	DQ	SQ	PQ
HoVer-Net	0.849	0.556	0.687	0.772	0.532	0.818	0.651	0.757	0.783	0.609
HoVer-Net+Res	0.849	0.558	0.685	0.774	0.531	0.823	0.650	0.765	0.813	0.631
HoVer-Net+Des	0.848	0.556	0.689	0.773	0.532	0.823	0.652	0.767	0.812	0.662
HoVer-Net+Res+Des	0.849	0.557	0.690	0.772	0.533	0.824	0.653	0.768	0.811	0.633
TSHVNet (Res)	0.847	0.554	0.680	0.776	0.528	0.825	0.660	0.776	0.813	0.633
TSHVNet (Des)	0.848	0.556	0.683	0.776	0.530	0.830	0.661	0.777	0.810	0.629
TSHVNet	0.856	0.558	0.690	0.777	0.546	0.835	0.662	0.779	0.813	0.637