Automatic detection of breast cancer for mastectomy based on MRI images using Mask R-CNN and Detectron2 models

Breast tumor diagnosis has seen widespread use of computer-aided techniques. Machine learning techniques can benefit doctors in making diagnosis decisions. One of the most important treatments for breast cancer is neoadjuvant chemotherapy (NAC). The reason is that NAC before surgery can downstage breast cancer and reduce local surgery. The problem of MRI, in brief, is how to distinguish between the types of pre-NAC and post-NAC, especially between the kinds of post-NAC. This study presents creating a system that goes through five stages: the input dataset, comparing normal and abnormal using EfficientNetV2L, determining the difference between malignant (pre- or post-NAC) and benign by utilizing a mask region-based convolutional neural network (R-CNN), comparing the types of post-NAC by using Detectron2, and finally the multidisciplinary team (MDT). Thus, it is decided if the breast needs a mastectomy or wide local excision (WLE) using Detectron2 with Faster R-CNN. The results showed that EfficientNetV2L achieved high accuracy, about 98%. The models successfully compared the types of post-NAC by using Detectron2 with Mask R-CNN. The study concludes that Detectron2 with Mask and Faster R-CNN is a reasonable model for detecting the type of MRI image and classifying whether the image is normal or abnormal.