Breast cancer is the most prevalent kind of cancer among women. Conventional techniques frequently depend on radiologists manually interpreting ultrasound pictures, which may be laborious and prone to human error-related accuracy fluctuation. This study presents an integrated deep learning approach for ultrasound picture-based breast cancer detection. A U-Net model is employed for precise tumor segmentation, followed by a Resnet, AlexNet and DenseNet model as a classifier to determine malignancy. Through early diagnosis and better patient care in the realm of medical imaging, this strategy seeks to increase the efficacy and accuracy of breast cancer detection. Convolutional neural networks (CNNs) are among the methods most often employed in image processing for deep learning. The results showed that the achieved high accuracy, about 98% for DenseNet algorithm among the three CNN Algorithms. This research showcases the potential of combining segmentation and classification techniques in a clinical context, offering a valuable tool for healthcare professionals.