Integrating Unet Segmentation and Vision Transformers for Improved Bone Cancer Diagnosis in Histopathological Images

Bone cancer represent a diagnosis challenge since this form of cancer appears rarely in patient populations. Traditional convolutional neural networks known as CNNs are frequently used for cancer detection and classification yet experience limited success within histopathological analysis tasks. The proposed work develops an innovative detection framework by integrating Unet segmentation algorithms alongside Vision Transformer (ViT) for better identification of bone cancer in histopathological image datasets. Diagnosis performance improves through our method because Unet delivers precise segmentation capabilities in addition to ViT's strengths in understanding global context. We boost bone cancer detection efficiency alongside reliability through data augmentation methods.