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.
- Immunocto (H&E)
- Acute Promyelocytic Leukemia (APL) (PBS)
We used an ensemble model consisting of ResNet18 and a ViT Large-16 adopted from
base architecture in torchvision.models. Training was done over
The final report is available here.