Hybrid segmentation and 3D Imaging: Comprehensive framework for breast cancer patient segmentation and classification based on digital breast tomosynthesis

Breast cancer is among the deadliest conditions and the second-leading reason for cancer mortality in women. Breast cells begin to develop into malignant, cancerous tumors, which is how breast cancer develops. Self-tests and routine professional examinations aid in early diagnosis, which increases the likelihood of surviving. A major difficulty for researchers is the categorization of breast cancer as a medical procedure. The prevailing aim of this research is the earlier recognition of breast cancer using 3D images and AI monitors. Initially, the mammogram images are collected and preprocessed to enhance the image quality. Noise removal is done using a Multistage Selective Convolution Filter (MSCF), contrast enhancement is done using an Enhanced Contrast Limited Adaptive Histogram (ECLAH), and an Improved Canny Operator (ICO) is employed for edge detection. Segmentation is done using the DENSE Squeeze Excitation Network (DENSE SE- Net). The preprocessed images are converted to 3D images using Digital Breast Tomosynthesis (DBT). From the hybrid segmented and 3D images, the objects are identified utilizing the You Only Look Once v7 (YOLO v7), and the images are classified into mild, medium, severe, and unaffected women using semi-supervised CNN (SS CNN). Finally, an AI monitor is used to monitor the patients. The simulation findings reveal that the suggested technique yields higher accuracy, precision, recall, and F1 score and better ROC characteristics than the existing approaches.