Medical imaging · computer vision
Brain tumor classification
Image-study classification with a hybrid CNN–Transformer architecture.
Problem
Separating classes in brain imagery needs both local detail (texture, edges) and global slice context. A pure CNN or a pure transformer usually falls short on one of the two.
Approach
A hybrid CNN–Transformer: the CNN supplies local spatial bias; the transformer models longer-range relations. Training and evaluation follow medical-vision research protocols.
Outcome
A research model for brain-image classification. Any accuracy figure depends on the set and the protocol; we do not fabricate them here.
Stack
PyTorch · CNN · Transformers · medical imaging