Extended Data Fig. 4: Benchmarking different segmentation methods with an admixture probability score.
From: Impact and correction of segmentation errors in spatial transcriptomics
a, Admixture probability scores for the mouse ileum dataset using three different segmentations (Baysor, ComSeg, and ProSeg). The scores for each cell type are shown as the trimmed mean scores relative to the original Baysor segmentation. We also show the scores for our cleaning procedure applied to each different segmentation. Box plot centerlines represent the median values, limits are the first and third quartiles, and whiskers end at the limits ±1.5 times the interquartile range. b, Admixture probability scores for the ovarian cancer dataset using four different segmentations (Xenium, Baysor, BIDCell, and ProSeg). The scores for each cell type are shown as the trimmed mean scores relative to the original Xenium segmentation. We also show the scores for our cleaning procedure applied to each different segmentation. Box plots are constructed the same way as in a.
