Self-supervised Learning for Cell Segmentation and Quantification in Digital Pathology Images

Fatemeh Haghighi, soumitra ghosh, Hai Ngu, Mohsen Hejrati, Sarah Chu, Han Lin, Somaye Hashemifar

Keywords: Medical Segmentation, Self-supervised Learning, contrastive learning

Abstract: Parkinson’s Disease (PD) is the second most common neurodegenerative disease in humans, impacting 2-3% of people over the age of 65. PD is characterized by the gradual loss of dopaminergic neurons in the Substantia Nigra (a part of the midbrain). At present, the number of dopaminergic neurons in the Substantia Nigra is one of the most important indexes in evaluating drug efficacy in PD animal models. Currently, analyzing and quantifying dopaminergic neurons is con- ducted manually by expert biologists through careful analysis of digital pathology images. However, this approach is laborious, time-consuming, and highly subjective, which significantly delays the progress in PD research and drug development. As such, a reliable and unbiased automated system is highly demanded for the quantification of neurons in digital pathology images. To this end, in this paper, we propose an end-to-end deep learning framework for the segmentation and quantification of dopaminergic neurons in PD animal models. Our framework relies on self-supervised learning advances to handle the limited amount of data for training deep models. Extensive experiments demonstrate the effectiveness of the developed method in quantifying neurons with a small amount of labeled data. As a result, the proposed methodology can lead to reliable data support for PD research and drug discovery by accelerating the digital pathology analysis.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Machine Learning for Sciences (eg biology, physics, health sciences, social sciences, climate/sustainability )

TL;DR: Identifying cell bodies in brain tissue digital pathology images