Self-supervised Learning of Inter-label Geometric Relationships for Gleason Grade Segmentation

Segmentation of Prostate Cancer (PCa) tissues from Gleason graded histopathology images is vital for accurate diagnosis. Although deep learning (DL) based segmentation methods achieve state-of-the-art accuracy, they rely on large datasets with manual annotations. We propose a method to synthesize PCa histopathology images by learning the geometrical relationship between different disease labels using self-supervised learning. Manual segmentation maps from the training set are used to train a Shape Restoration Network (ShaRe-Net) that predicts missing mask segments in a self-supervised manner. Using DenseUNet as the backbone generator architecture we incorporate latent variable sampling to inject diversity in the image generation process and thus improve robustness. Experimental results demonstrate the superiority of our method over competing image synthesis methods for segmentation tasks. Ablation studies show the benefits of integrating geometry and diversity in generating high-quality images. Our self-supervised approach with limited class-labeled data achieves better performance than fully supervised learning.

Automated summary

Our proposed method synthesizes PCa histopathology images using self-supervised learning, outperforming competing methods and achieving better performance in segmentation tasks.