I worked as a research assistant in computational histopathology from January to May 2025.

Background

Pathologists are medical specialists that know how to interpret human tissue in microscopes. They are able to tell if there is one disease in the tissue sample, so histopathology is the science and practice of diagnosing from tissue samples. As with the entire world, there is a shortage of medical professionals, so developing computational techniques for that is interesting.

My contribution

My project was to train and compare segmentation models using different modalities of input data. For that, we used a dataset from our collaborators in UEF/Kuopio. Our dataset has tissue samples either Chemically Stained, Unstained, Virtually Stained and a binary mask telling which regions of this tissue are gland and which are not. I used a basic unet and an attention unet as my models, and I used all of the available images to train those models.