Zebrafish CHT Cell Counter
Pretrained U-Net models for automated analysis of zebrafish fluorescence microscopy images. The pipeline segments the caudal hematopoietic tissue (CHT) region, detects hematopoietic stem and progenitor cells (HSPCs), and estimates cell counts within the CHT while providing pixel-wise masks.
Available model families
- Keyence: Models trained on images acquired with a Keyence BZ-X710 fluorescence microscope, centered on the CHT region of the zebrafish embryo.
- Chem-screen: Models trained on chemical-screening microscopy images acquired on PerkinElmer Opera Phenix Plus High-Content Screening System.
- Combined: An HSPC model trained on both datasets, paired with size-dependent CHT segmentation models.
Usage
These checkpoints are designed for use with the zf_cht_cell_counter Python package, which handles model downloads, preprocessing, inference, and count exports.
This repository contains model checkpoints, training metadata, and geometry configurations. The imaging datasets can be found at [Zenodo DOI](xavier add link here).
Intended use and limitations
Performance varies by image resolution, fluorescence intensity, acquisition conditions, and biological differences. Please inspect predicted masks and validate cell counts against manual annotations before analyzing a new dataset.