Unsupervised Learning of Cell Instances with Generative Routing Pyramids
2026-08-17 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionMachine Learning
AI summaryⓘ
The authors developed a new way to find and describe individual cells in microscope images without needing any labeled examples. Their method works by breaking down images into parts linked to hidden features that represent each cell's shape and position. This helps separate each cell (instance segmentation) and also captures what the cell looks like (phenotypic classification). They tested their method on various types of cell images and found it works well, even creating new cell images based on changes. The authors also shared their code online for others to use.
instance segmentationphenotypic classificationmicroscopy image analysisunsupervised learninglatent variablesimage reconstructioncell morphologymachine learninggenerative modeling
Authors
Ziwen Liu, Martin Weigert
Abstract
Identifying and representing object instances such as cells or nuclei is a common task in microscopy image analysis. Established machine learning workflows typically use supervised detection or segmentation followed by feature extraction or classification, which requires manual annotations and treats instance segmentation and cell representation as separate stages. We describe a new unsupervised method for cell instance segmentation and phenotypic classification from unlabeled microscopy images. Our method is based on reconstructing each image using a coarse-to-fine routing pyramid that associates pixels with spatially sparse latent sources. The resulting pixel-to-latent associations yield instance masks, while the source latents encode cell morphology. We demonstrate competitive performance in instance segmentation across diverse cell morphologies and imaging modalities, as well as generative modeling of cellular phenotypes under perturbations. Source code and checkpoints are available at https://github.com/weigertlab/routing-pyramids.