Rotation-Agnostic Image Representation Learning for Digital Pathology
Saghir Alfasly, Abubakr Shafique, Peyman Nejat, Jibran Khan, Areej Alsaafin, Ghazal Alabtah, Hamid R. Tizhoosh
Abstract
This paper addresses complex challenges in histopatho-logical image analysis through three key contributions. Firstly, it introduces a fast patch selection method, FPS, for whole-slide image (WSI) analysis, significantly reducing computational cost while maintaining accuracy. Sec-ondly, it presents PathDino, a lightweight histopathol-ogy feature extractor with a minimal configuration of five Transformer blocks and only <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\approx 9$</tex> million parameters, markedly fewer than alternatives. Thirdly, it introduces a rotation-agnostic representation learning paradigm using self-supervised learning, effectively mitigating overfuting. We also show that our compact model outperforms existing state-of-the-art histopathology-specific vision transformers on 12 diverse datasets, including both internal datasets spanning four sites (breast, liver, skin, and colorectal) and seven public datasets (PANDA, CAMELYON16, BRACS, DigestPath, Kather, PanNuke, and WSSS4LUAD). Notably, even with a training dataset of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\approx 6$</tex> million histopathol-ogy patches from The Cancer Genome Atlas (TCGA), our approach demonstrates an average 8.5% improvement in patch-level majority vote performance. These contributions provide a robust framework for enhancing image analysis in digital pathology, rigorously validated through extensive evaluation.<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup><sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup>The project page: https://KimiaLabMayo.github.io/PathDino-Page/