Search for a command to run...
Breast cancer is the most serious life-threatening tumours in women globally, affecting millions each year. Early detection, influenced by genetics, lifestyle, and environment, is critical through approaches like as mammography. This paper presents a method for segmenting region of interest (ROI) in mammography images using a computer-aided detection (CAD) model. The model is trained and tested using the Mini-DDSM dataset. Given that the dataset contains unclear and low-quality images, it addresses this issue by denoising and segmenting ROIs with medium filtration and applying the CLAHE technique to the dataset. Similarly, the CLAHE significantly brightens a mammographic image, enabling more precise delineation of noncancerous and cancerous regions, and it also improves the dice coefficient, IoU, which helps to segment the tumours correctly. Breast cancer is diagnosed and segmented using deep learning methodologies, namely U-net architecture based on the pre trained VGG19 backbone segmentation methods. A U-Net architecture based on the VGG19, which extracts relevant features from mammography images this is used to create an advanced model for segmenting regions of interest in mammograms. The model can effectively extract relevant features from mammographs using pretrained VGG19 layers. Custom data generators provide seamless batch processing during model training, while critical metrics like as dice coefficient, intersection over union (IOU), recall, and precision are monitored to ensure the model's performance is optimized. This analyses the model's accuracy (96.7%) in segmenting areas of breast cancer using a test dataset, offering informative visual comparisons