An improved hybrid feature selection and classification framework for breast cancer detection using mammography images

Breast cancer is one of the most common diseases that affects women around the world. Finding it early can help save lives. Doctors usually use tests like biopsy, ultrasound, CT scan, and mammography to detect breast cancer. In this study, we created a computer-based method that helps in detecting breast cancer more accurately. The process has four main steps — preprocessing, segmentation, feature selection, and classification. First, we clean the mammogram images using a filter to remove unwanted noise. Next, we separate the important part of the image using a special method called Thresholding-Based Level Set. Then, we choose only the most important details (features) from the image using a new hybrid method called Improved Grey Wolf Optimization with Seagull Optimization Algorithm. Finally, we use a machine learning model named CatBoost to identify if the tumor is benign (non-cancerous) or malignant (cancerous). When tested on a dataset, our method showed excellent results with 99.2 % accuracy. This shows that our model can help doctors detect breast cancer early and more correctly.