Advancing Breast Cancer Detection: Enhancing YOLOv5 Network for Accurate Classification in Mammogram Images

Recent advances in artificial intelligence (AI), notably deep learning, have sparked widespread curiosity with bioinformatics, particularly the challenges presented by medical imaging. It's been really helpful in enabling the Computer Aided Diagnosis CAD system to provide precise outcomes. Nonetheless, it is still a difficult task to identify breast cancer in mammography images. The purpose of this effort is to lower False Positive Rate FPR and False Negative Rate FNR and increase Matthews's correlation coefficient MCC value. Two highly tailored object detection models, YOLOv5 and Mask R-CNN, are utilized to get the job done. YOLOv5 is able to detect the mass and determine whether it is benign or malignant. However, YOLOV5's limited real estate necessitates certain tweaks to the original model in order to get the desired effects. Tumor borders and size are both identified by Mask RCNN as it traverses breast parenchyma in search of malignancies. Stages of cancer are based on the magnitude of the patients' tumours. This model employs YOLOv5+Mask RCNN and is trained on the INbreast, CBIS-DDSM, and BNS dataset. The proposed model is compared against the baseline version of YOLOv5 to determine how well it performs. The proposed method improves performance, with an FPR of 0.049%, a FNR of 0.029%, and a high MCC value of 92.02%. Based on the results of the studies, combining YOLOv5 with Mask RCNN improves accuracy by 0.06 percentage points compared to using either method alone. Furthermore, this effort may aid in determining the patient's prognosis and allowing clinicians to be more accurate and predictable in the diagnosing process at an early stage.