Table 3. Performance benchmarking with...

Table 3.

Performance benchmarking with self, weakly and fully supervised methods in the MoNuSeg data set

Method	AJI	F1	Dice
Self supervised
DARCNN (Hsu et al., 2021)	0.446	0.5410	–
Self-Attention (Sahasrabudhe et al., 2020)	0.535	–	0.747
CyC-PDAM (Liu et al., 2020)	0.561	0.748	–
Nucleus-Aware (Song et al., 2023)	0.593	0.759	–
Weakly supervised
Partial points (Qu et al., 2019)	0.543	0.776	0.732
Point annotations (Lin et al., 2023)	0.562	0.776	0.744
BoNuS (Lin et al., 2024)	0.607	0.780	0.767
Cyclic learning (Zhou et al., 2023)	0.636	0.774	0.774
Fully supervised
U-Net (Ronneberger et al., 2015)	0.543	0.779	–
RCSAU-Net (Wang et al., 2022)	0.619	0.82	–
HoVer-Net (Graham et al., 2019)	0.618	0.826	–
CDNet (He et al., 2021)	0.633	0.831	–
TopoSeg (He et al., 2023)	0.643	–	–
GUSL (Yang et al., 2025)	0.673	0.886	0.803
NucleiSegNet (Lal et al., 2021)	0.688	0.813	–
CIA-Net (Zhou et al., 2019)	0.691	0.901	–
LG-NuSegHop (baseline)	0.651	0.887	0.778
LG-NuSegHop (dom. Adapted)	0.658	0.892	0.791

Method	AJI	F1	Dice
Self supervised
DARCNN (Hsu et al., 2021)	0.446	0.5410	–
Self-Attention (Sahasrabudhe et al., 2020)	0.535	–	0.747
CyC-PDAM (Liu et al., 2020)	0.561	0.748	–
Nucleus-Aware (Song et al., 2023)	0.593	0.759	–
Weakly supervised
Partial points (Qu et al., 2019)	0.543	0.776	0.732
Point annotations (Lin et al., 2023)	0.562	0.776	0.744
BoNuS (Lin et al., 2024)	0.607	0.780	0.767
Cyclic learning (Zhou et al., 2023)	0.636	0.774	0.774
Fully supervised
U-Net (Ronneberger et al., 2015)	0.543	0.779	–
RCSAU-Net (Wang et al., 2022)	0.619	0.82	–
HoVer-Net (Graham et al., 2019)	0.618	0.826	–
CDNet (He et al., 2021)	0.633	0.831	–
TopoSeg (He et al., 2023)	0.643	–	–
GUSL (Yang et al., 2025)	0.673	0.886	0.803
NucleiSegNet (Lal et al., 2021)	0.688	0.813	–
CIA-Net (Zhou et al., 2019)	0.691	0.901	–
LG-NuSegHop (baseline)	0.651	0.887	0.778
LG-NuSegHop (dom. Adapted)	0.658	0.892	0.791

Sharing Unavailable