ikeda042/phenopixel

PhenoPixel is a web application built for scientists who work with microscopy images of individual cells, particularly bacteria. It is deployed at Hiroshima University and handles a specific file format called ND2, which is produced by Nikon microscopes. You upload one of those files, the tool finds the outlines of individual cells in the images, and then lets you inspect, label, and measure those cells in detail. The application has a Python backend (FastAPI) and a React frontend. On the backend, it uses a computer-vision library called OpenCV to detect cell contours in microscope images. Each detected cell gets stored in a local SQLite database, and from there you can view the cell in several ways: as a plain outline, as a fluorescence overlay, as a heat map of signal intensity along the cell's length, or as a 256-level intensity map useful for studying where things are located inside the cell. Fluorescence imaging is a lab technique that makes specific molecules glow, and PhenoPixel can handle images with up to four separate fluorescence channels. A key part of the workflow is annotation. Automated cell detection sometimes picks up debris or cells that are stuck together. The annotation screen lets you manually review what was detected and move items into labeled groups before running bulk measurements. Once a group is clean, the bulk engine can calculate cell length, cell area, fluorescence intensity summaries, and other descriptors across the whole population, then export the results as CSV or JSON. Setting it up requires Python and Node.js running in separate terminals. The repository also includes Docker support and a documentation site built with Docusaurus. The project is aimed at researchers running single-cell phenotype studies and provides enough export options to feed the results into standard analysis pipelines.