Magnification Prior: A Self-Supervised Method for Learning Representations on Breast Cancer Histopathological Images

This work presents a novel self-supervised pre-training method to learnefficient representations without labels on histopathology medical imagesutilizing magnification factors. Other state-of-theart works mainly focus onfully supervised learning approaches that rely heavily on human annotations.However, the scarcity of labeled and unlabeled data is a long-standingchallenge in histopathology. Currently, representation learning without labelsremains unexplored for the histopathology domain. The proposed method,Magnification Prior Contrastive Similarity (MPCS), enables self-supervisedlearning of representations without labels on small-scale breast cancer datasetBreakHis by exploiting magnification factor, inductive transfer, and reducinghuman prior. The proposed method matches fully supervised learningstate-of-the-art performance in malignancy classification when only 20% oflabels are used in fine-tuning and outperform previous works in fullysupervised learning settings. It formulates a hypothesis and provides empiricalevidence to support that reducing human-prior leads to efficient representationlearning in self-supervision. The implementation of this work is availableonline on GitHub -https://github.com/prakashchhipa/Magnification-Prior-Self-Supervised-Method