Image segmentation methods for automated morphological analysis of organoids

Organoids — three dimensional, self-organized cell cultures derived from stem cells — are a powerful platform for the study of diseases and their possible treatments. Organoids are typically created by first growing spheroids — three dimensional cultures of only a single cell type — of induced pluripotent stem cells (iPSCs).

In a separate work, we are studying the early growth of several hundred of those cultures, resulting in tens of thousands of microscopy images. We are using computer vision and machine learning techniques to automatically measure the size and shape of the developing spheroids. Many tools require a significant amount of task-specific training data, but recent advances in “0-shot” methods allow for image segmentation without any specific retraining.

Here we report on the performance of several of these tools by testing them on a subset of the data and comparing with the results of manual image segmentation. We test a variety of image analysis techniques, some general purpose, some specifically developed for microscopy images. We find that combining multiple tools can dramatically improve the results both for methods requiring retraining and for methods not requiring it.