Immunostain Image Analysis

This image analysis program was conceived as a means to streamline the tasks of de-noising and emphasizing regions of interest (ROIs) in images acquired via immunofluorescence imaging. Its primary function is to take a user-provided red-green-blue (RGB) fluorescence image or image folder input, isolate a specific stain channel to be examined, optimize the image(s) for analysis through a series of processes including noise filtration and selective exclusion of non-ROIs, and then collect and compile data points of interest from the post-processing images, from which tangible and, ideally, reliable results can be drawn.

Capabilities

At the moment, the program works very reliably when provided with the requisite user inputs, including an image file pathway, the output file location, the stain channel one wishes to create a mask of, and the nucleus size threshold defining the minimum pixel cluster area for which objects - in this case cells - will be included in the data analysis.

Methodology/Algorithm(s)

In its current form, this program consists of an entirely Python-based backbone for its interface, object management, and practical applicability of the algorithm(s) used. The code was written in both Miniconda and Pycharm environments, with a variety of different function packages being utilized as outlined in the separate requirements.txt file.

The actual image analysis is performed by a U-Net neural network, which has been trained on immunofluorescence images to be able to accurately identify ROIs (in this case mesenchymal stem cells) and isolate them from undesired objects (often debris or other non-stem cells mixed in with the experimental batch). Those defined ROIs then become the areas from which various data is taken - namely stain pixel intensity and distribution - and quantified numerically in the form of an Excel spreadsheet.