Breast Cancer Segmentation and Detection Using U-net With Ultrasound Dataset
Md. Mijnaur Rahman, Anwar Hossain, Nipu Chakraborty, Md Rezaul Karim Khan · 2024
Breast cancer is a significant health issue for women, being one of the most common types of cancer and the second leading cause of death among them. Detecting breast cancer early is crucial for effective treatment. While ultrasound is often used for screening, diagnosing cancer manually takes a lot of time and is not always accurate. This study introduces a new way to find breast cancer in ultrasound images using a deep-learning model called U-net. This study used techniques like binary cross-entropy and Adam optimization to make the method strong and reliable. The BCE Dice Loss, which combines binary cross-entropy and the Dice Coefficient, helps the authors precisely locate the cancer. This method isn't just effective for ultrasound images – it can also work with mammography and MRI images. Plus, it's not limited to breast cancer; it can help detect other medical issues, too. When the researchers tested this method with ultrasound images, they found it had a training accuracy of 99.37% and a test accuracy of 96.2%, showing how well it performs.