Segmentation-Based Classification Deep Learning Model for Breast Cancer Detection using Mammogram images

Breast cancer has emerged as a leading cause of mortality, responsible for an extensive number of deaths in recent years. The current imaging-based diagnostic methods adopted for the detection of breast cancer, such as mammography, has shown inadequate effectiveness in clinical environments due to their tendency for significant mistake rates. This paper introduces an effective methodology that uses segmentation based on deep learning classifiers for classification in order to perform an automated, productive, and precise diagnosis of breast cancer. In order to enhance the best model combination, a hybrid approach was employed, integrating deep learning segmentation models with VGG-19 classification models. The performance of the proposed methodology was evaluated using several statistical metrics, such as accuracy, precision, recall, f1-score, and receiver operating characteristics (ROC), along with cross-entropy loss function. The proposed methodology showed outstanding results compared to other segmentation-based classification model techniques. The research concluded that the UNet segmentation method, along with an improved VGG-19 classifier, had an improvement of 2.25% as compared to the VGG-19 (Pre-Trained) model.