Breast Cancer detection using Transfer Learning

Globally, breast cancer continues to be the most prevalent and severe type of cancer among women. For better survival rates and efficient treatment, early and precise detection is essential. In this work, we introduce a deep learning based method that uses the Resnet50 model and Transfer Learning to detect and provides the classification score of breast cancer ultrasound images. To enhance image efficiency and model generalization, the suggested system uses sophisticated pre processing methods such data augmentation, normalization, and noise reduction. To speed up training and increase accuracy, a pre trained model is adjusted using domain specific ultrasound image data. Experiments on the Breast Ultrasound Image Dataset (BUSI) show encouraging results; the classification model maintains low loss values while attaining an accuracy of 95.33% and validation accuracy of 90.15%. With time, the segmentation component’s detection ability improves when it is tuned using a dice loss function. When utilized in a desktop application, this integrated technique greatly improves automated breast cancer diagnostics speed and reliability. The findings highlight Transfer Learning’s potential for early breast cancer diagnosis in practical clinical settings.