A periodic health check with physical examination and assessment of the current health status is among the best approaches to protect your health. Through that, we can know if we have any health problems and promptly offer a treatment plan if the problems are detected. In most cases, early diagnosis and disease detection are very important, especially for cancer. Breast cancer is serious and also a very common disease in women. However, if the disease is detected early, the chance of a cure is very high. Deep learning-based segmentation methods have been introduced to detect Breast cancer tumors, but we are facing challenges in the limitations of data. Although some data augmentation approaches have been presented, the number of augmented samples should be further considered. This work has examined the efficiency of data augmentation techniques by Brightness and Rotation with various ratios of increased samples on ultrasound images. Data augmentation improves the performance using U-NET to perform segmentation tasks for breast cancer diagnosis. The experimental results show that the rotation technique can increase the average performance in the training and test phases.