Efficient Self-Supervised Grading of Prostate Cancer Pathology
Riddhasree Bhattacharyya, Surochita Pal Das, Sushmita Mitra
TL;DR
This paper tackles ISUP grading of prostate cancer from WSIs using only slide-level labels, addressing patch-level annotation scarcity, gigapixel image complexity, and staining variability. It introduces TSOR, a three-module framework that (i) builds a balanced patch-level SSL dataset via MIL, (ii) learns task-specific and stain-agnostic features through a teacher-student SSL with dual augmentations and a stain-augmentation loss, and (iii) fine-tunes with ordinal regression using an attention-based MIL head for ISUP grading. On PANDA and SICAP datasets, TSOR achieves state-of-the-art performance, including a PANDA Kappa of about 0.884 and high AUC/F1 across tasks, while demonstrating robustness to stain variation and offering interpretable patch-level explanations. By reducing reliance on patch-level annotations and providing interpretable results, TSOR has potential to streamline clinical workflows and improve consistency in prostate cancer grading across centers.
Abstract
Prostate cancer grading using the ISUP system (International Society of Urological Pathology) for treatment decisions is highly subjective and requires considerable expertise. Despite advances in computer-aided diagnosis systems, few have handled efficient ISUP grading on Whole Slide Images (WSIs) of prostate biopsies based only on slide-level labels. Some of the general challenges include managing gigapixel WSIs, obtaining patch-level annotations, and dealing with stain variability across centers. One of the main task-specific challenges faced by deep learning in ISUP grading, is the learning of patch-level features of Gleason patterns (GPs) based only on their slide labels. In this scenario, an efficient framework for ISUP grading is developed. The proposed TSOR is based on a novel Task-specific Self-supervised learning (SSL) model, which is fine-tuned using Ordinal Regression. Since the diversity of training samples plays a crucial role in SSL, a patch-level dataset is created to be relatively balanced w.r.t. the Gleason grades (GGs). This balanced dataset is used for pre-training, so that the model can effectively learn stain-agnostic features of the GP for better generalization. In medical image grading, it is desirable that misclassifications be as close as possible to the actual grade. From this perspective, the model is then fine-tuned for the task of ISUP grading using an ordinal regression-based approach. Experimental results on the most extensive multicenter prostate biopsies dataset (PANDA challenge), as well as the SICAP dataset, demonstrate the effectiveness of this novel framework compared to state-of-the-art methods.