Improved Bone Cancer Diagnosis: Transfer Learning Integration with U-Net for Segmented Histopathology Image Analysis

Bone cancer, while rare, poses significant health challenges with often late diagnoses. Though CNN, proposed asa good algorithm for classifying and detecting cancer from images, it faces challenges when it is a histopathological image. This study introduces a novel approach combining machine learning with cutting-edge image processing methods to increase the precision and effectiveness of diagnosis. The model retains several key features. Utilizing U-Net for segmentation and four different CNNs, ResNet50, VGG16, MobileNetV2, andDenseNet121 for classification, along with data augmentation strategies, our method aims to enhance the reliability of bone cancer detection. Preliminary results demonstrate improved performance in accuracy and speed compared to traditional methods.