Fig. 2: HPC analysis and pathologist validation of mesothelioma tissue patterns. | Nature Communications

Fig. 2: HPC analysis and pathologist validation of mesothelioma tissue patterns.

From: A histomorphological atlas of resected mesothelioma discovered by self-supervised learning from 3446 whole-slide images

Fig. 2

a A HPC color-coded UMAP (Uniform Manifold Approximation and Projection) plot displaying the spatial distribution of 47 distinct HPCs. b Distribution and patient prevalence of HPCs: Percentage of LATTICe-M cases exhibiting each HPC, with colouring indicating rare and ubiquitous (<20% or  >80%, grey) and intermediate (20−80%, blue) patterns. Also on the right, distribution of HPCs categorised by patient prevalence decile. c Majority consensus for pathologists' panel’s annotations for HPCs. Asterisks (*) indicate complete agreement among all pathologists. d Main pattern color-coded UMAP, where annotations were provided by a pathologist following the clustering step to group HPCs with similar morphologies. e WSI from a poor outcome and good outcome case overlaid and quantified by HPCs, demonstrating the differences in HPC composition and showing the percentage of the top 10 HPCs for each sample, alongside their histomorphology annotations (scale bar, 21mm). Source data are provided as a Source Data file.

Back to article page