Breast cancer is a leading cause of mortality among women worldwide, highlighting the urgent need for accurate, accessible, and non-invasive diagnostic solutions. Ultrasound imaging, due to its cost-effectiveness and safety, is widely used in breast cancer detection. This study introduces a deep learning-based approach for classifying breast tumors in ultrasound images through a pipeline combining segmentation and feature extraction. Preprocessing techniques are first applied to reduce noise and enhance image clarity. Tumor regions are then segmented using the TransUNet architecture, which effectively integrates convolutional and transformer-based mechanisms. MobileNetV2 is subsequently employed for feature extraction, leveraging its lightweight and efficient architecture. The extracted features are used to train classification models to distinguish between benign and malignant breast masses. This integrated methodology shows strong potential in assisting radiologists with accurate decision-making and early diagnosis. Future work will focus on validating the approach using larger and more diverse datasets to ensure robustness and generalizability.

Keywords:

Breast Cancer, Ultrasound Imaging, Deep Learning, TransUNet, MobileNetV2, Image Segmentation, Feature Extraction, Tumor Classification, Medical Image Analysis, Computer-Aided Diagnosis.