This study proposes a deep learning approach for early breast cancer detection using the BUSI dataset. Enhanced U-Net models perform tumor segmentation, while InceptionV3 and ResNet50 classify tumors as benign or malignant. The method improves detection accuracy and supports automated clinical diagnosis.