Skip to main content
Fig. 4 | BMC Bioinformatics

Fig. 4

From: CellSeg: a robust, pre-trained nucleus segmentation and pixel quantification software for highly multiplexed fluorescence images

Fig. 4

CellSeg performs comparably to established deep learning-based segmentation algorithms on diverse human FFPE tissues. Representative images from tissues described in Fig. 3 are shown. A. StarDist, Cellpose, and CellSeg show comparable performance on spleen. B. StarDist oversegments several spindly nuclei in DFSP (arrows), while CellSeg and Cellpose segment nuclei accurately. C. StarDist and CellSeg segment more low intensity objects in GBM (arrows). D. all three algorithms perform similarly well on HCC. E. StarDist and CellSeg segment more low intensity objects in seminoma (arrows). F. All three algorithms perform relatively poorly on T-ALL. Scale bar, 20 μm

Back to article page