Skip to content

hoseok-lee/immune-classifier

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

19 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Objective

Binary classification of histopathology images cells as immune or non-immune based on learned morphological features. We performed cross-validation across two imaging domains, hematoxylin & eosin (H&E) staining and peripheral blood smears (PBS), to investigate whether learned features are transferable.

Datasets

  1. Immunocto (H&E)
  2. Acute Promyelocytic Leukemia (APL) (PBS)

Model

We used an ensemble model consisting of ResNet18 and a ViT Large-16 adopted from base architecture in torchvision.models. Training was done over $20$ epochs on GPUs at Digital Research Alliance of Canada (DRAC). UNI and DinoBloom were trained as baseline models for state-of-the-art (SOTA) comparison. This directory contains code for loading the two datasets and training all models discussed.

The final report is available here.

About

Investigation of immune cell classification performance of deep-learning models across different microscopy imaging domains.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages