Histology-informed tiling of whole tissue sections improves the interpretability and predictability of cancer relapse and genetic alterations

TL;DR

This paper introduces histology-informed tiling (HIT), a method that uses tissue architecture to improve the interpretability and accuracy of models predicting cancer relapse and genetic alterations from whole slide images.

Contribution

HIT leverages semantic segmentation to extract biologically meaningful tissue structures, enhancing model performance and interpretability in digital pathology.

Findings

HIT achieved a gland-level Dice score of 0.83.

Improved MIL models' AUCs by 10% for key genetic alterations.

Identified gland clusters associated with relapse and mutations.

Abstract

Histopathologists establish cancer grade by assessing histological structures, such as glands in prostate cancer. Yet, digital pathology pipelines often rely on grid-based tiling that ignores tissue architecture. This introduces irrelevant information and limits interpretability. We introduce histology-informed tiling (HIT), which uses semantic segmentation to extract glands from whole slide images (WSIs) as biologically meaningful input patches for multiple-instance learning (MIL) and phenotyping. Trained on 137 samples from the ProMPT cohort, HIT achieved a gland-level Dice score of 0.83 +/- 0.17. By extracting 380,000 glands from 760 WSIs across ICGC-C and TCGA-PRAD cohorts, HIT improved MIL models AUCs by 10% for detecting copy number variation (CNVs) in genes related to epithelial-mesenchymal transitions (EMT) and MYC, and revealed 15 gland clusters, several of which were associated with cancer relapse, oncogenic mutations, and high Gleason. Therefore, HIT improved the accuracy and interpretability of MIL predictions, while streamlining computations by focussing on biologically meaningful structures during feature extraction.