Implementation of Breast Cancer Segmentation Using Improved Attention UNet

Breast cancer continues to be a serious global health concern around the world, demanding the development of advanced computer models to accomplish precise detection and segmentation of benign and malignant tissues. In this study, we address this demand by proposing a refined segmentation technique through using a fine-tuned Attention U-Net model. In order to boost the effectiveness of the model, we applied batch normalization between convolutional layers and integrated dropout layers in the encoder architecture to solve the issue of overfitting. The methodology we proposed provided outstanding outcomes, with an incredible pixel precision of 97.7%. Likewise with the accuracy score, we also obtained satisfactory Iou of 0.88 and strong recall and precision rates of 94% and 92%. These findings enable our model to distinguish between the masses of tumours in breasts and provide future avenues for improved diagnostic and treatment options.