A Self-Supervised Learning Approach for High Throughput and High Content Cell Segmentation

Abstract

In principle, AI-based algorithms should enable rapid and accurate cell segmentation in high-throughput settings. However, reliance on large datasets, human input, and computational expertise, along with issues of limited generalizability and the necessity for specialized training are notable drawbacks of nominally "automated" segmentation tools. To overcome this roadblock, we introduce an innovative, user-friendly self-supervised learning method (SSL) for pixel classification that requires no dataset-specific modifications or curated labelled data sets, thus providing a more streamlined cell segmentation approach for high-throughput and high-content research. We demonstrate that our algorithm meets the criteria of being fully automated with versatility across various magnifications, optical modalities and cell types. Moreover, our SSL algorithm is capable of identifying complex cellular structures and organelles which are otherwise easily missed, thereby broadening the machine learning applications to high-content imaging. Our SSL technique displayed consistent F1 scores across segmented images, with scores ranging from 0.831 to 0.876, outperforming the popular Cellpose algorithm, which showed greater variance in F1 scores from 0.645 to 0.8815, mainly due to errors in segmentation. On average, our SSL method achieved an F1 score of 0.852 {+/-}0.017, exceeding Cellposes average of 0.804 {+/-}0.08. This novel SSL method not only advances segmentation accuracy but also minimizes the need for extensive computational expertise and data security concerns, making it easier for biological researchers to incorporate automated segmentation into their studies.