Exploring Non-contrastive Self-supervised Representation Learning for Image-based Profiling

2025-06-17

AI Summary

Addressing the substantial distributional shift between cellular and natural images, as well as the difficulty of effectively fusing multi-view information, this work proposes an unsupervised feature learning framework tailored for cellular phenotypic analysis. We pioneer the adaptation of non-contrastive self-supervised learning (e.g., BYOL, SwAV) to this domain, introducing cell-specific geometric and photometric augmentation strategies, and designing a multi-view feature aggregation and normalization module to enable robust representation learning. The method operates entirely without labeled data and significantly improves cross-cell-line transferability. Evaluated on the CVPR 2025 Cell Line Transferability Challenge, it achieves first place. Our approach establishes a generalizable, plug-and-play feature extraction paradigm for image-based cellular analysis in drug discovery—overcoming limitations of supervised models trained on narrow, label-intensive datasets.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Computer Vision: Representation Learning for Vision
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Image-based cell profiling aims to create informative representations of cell images. This technique is critical in drug discovery and has greatly advanced with recent improvements in computer vision. Inspired by recent developments in non-contrastive Self-Supervised Learning (SSL), this paper provides an initial exploration into training a generalizable feature extractor for cell images using such methods. However, there are two major challenges: 1) There is a large difference between the distributions of cell images and natural images, causing the view-generation process in existing SSL methods to fail; and 2) Unlike typical scenarios where each representation is based on a single image, cell profiling often involves multiple input images, making it difficult to effectively combine all available information. To overcome these challenges, we propose SSLProfiler, a non-contrastive SSL framework specifically designed for cell profiling. We introduce specialized data augmentation and representation post-processing methods tailored to cell images, which effectively address the issues mentioned above and result in a robust feature extractor. With these improvements, SSLProfiler won the Cell Line Transferability challenge at CVPR 2025.

Problem

• Develop non-contrastive SSL for cell image representation learning
• Address distribution gaps between cell and natural images
• Handle multi-image inputs for effective cell profiling

Innovation

• Non-contrastive SSL for cell images
• Specialized data augmentation methods
• Representation post-processing techniques