HistoSSL: Self-Supervised Representation Learning for Classifying Histopathology Images

Abstract

The success of image classification depends on copious annotated images for training. Annotating histopathology images is costly and laborious. Although several successful self-supervised representation learning approaches have been introduced, they are still insufficient to consider the unique characteristics of histopathology images. In this work, we propose the novel histopathology-oriented self-supervised representation learning framework (HistoSSL) to efficiently extract representations from unlabeled histopathology images at three levels: global, cell, and stain. The model transfers remarkably to downstream tasks: colorectal tissue phenotyping on the NCTCRC dataset and breast cancer metastasis recognition on the CAMELYON16 dataset. HistoSSL achieved higher accuracies than state-of-the-art self-supervised learning approaches, which proved the robustness of the learned representations.

Contrastive Deep Encoding Enables Uncertainty-aware Machine-learning-assisted Histopathology

“…Interestingly, this model outperformed both the current SOTA, i.e. HistoSSL-Res [49], and the 100% baseline model.…”

“…For selected cases, MAE, Simclrv1, and Simclrv2 models were modified with uncertainty-aware loss; the corresponding results are shown in the "Uncertainty-aware Model" columns. Results marked by * are quoted from[36]; Results marked by * * are quoted from[49]. The rest are from our experiments.…”

A novel automatic annotation method for whole slide pathological images combined clustering and edge detection technique

“…This high workload can lead to visual fatigue, increasing the risk of misdiagnosis. Fortunately, with the rapid advancement of computer technology, obtaining digital pathological slide images has become easier [5]. This enables the use of computer‐aided diagnosis and treatment technology, which can assist pathologists in making basic diagnoses, improve the efficiency of pathological diagnosis, and thereby alleviate the shortage of pathologists and inadequate medical resources in the current healthcare landscape.…”

“…In recent years, the emergence of Whole Slide Images (WSIs) [5] has led to the widespread use of recognition methods that combine machine learning and deep learning in clinical histopathological image analysis. In the early stages of research, traditional image analysis or machine learning methods were commonly employed for classifying pathological image blocks.…”