Practical Considerations for Using InstanSeg v Cellpose in QuPath for Multiplex Immunofluorescence

Practical Considerations for Using InstanSeg vs. Cellpose in QuPath for Multiplex Immunofluorescence Analysis | Ved Sharma | November 18th, Halfway to I2K 2025 Authors: Ved Sharma, Bio-Imaging Resource Center, The Rockefeller University Workshop Description: Integration of deep learning segmentation methods (Cellpose, InstanSeg, and others) into QuPath makes it possible to analyze large 2D whole-slide images, comprising tens of thousands to millions of cells at single cell resolution with greater accuracy than classical segmentation methods. In this demo, I'll go over a multiplex immunofluorescence (mIF) analysis pipeline I put together for whole slide tumor tissue sections. The primary focus of the demo will be on the nucleus/cell segmentation part of the pipeline. I'll discuss practical issues, advantages, and limitations one should consider when deciding between InstanSeg and Cellpose to segment nuclei/cells in mIF images. Basic familiarity with the QuPath program, Cellpose and InstanSeg extensions is advantageous, but not required. QuPath: https://qupath.readthedocs.io/en/stable/ Cellpose extension: https://github.com/BIOP/qupath-extension-cellpose InstanSeg extension: https://qupath.readthedocs.io/en/stable/docs/deep/instanseg.html Target Audience: Beginner users, intermediate users Keywords: QuPath, Cellpose, InstanSeg, nucleus/cell segmentation, AI, deep learning