Enhancing Breast Cancer Detection through Multimodal 2D Convolutional Neural Networks: Integrating Mammogram Imaging and Numerical Features

Abstract

Breast cancer is a significant cause of mortality, highlighting the need for early and accurate detection to improve patient outcomes. Traditional mammography, though widely used, has limitations in distinguishing malignancies, particularly in cases of dense breast tissue, calcifications, and fatty breast. This study proposes a multimodal 2D Convolutional Neural Network (m2D-CNN) model that combines mammogram imaging with key numerical features—Mean, Radius, and Area—extracted from the Region of Interest (ROI) to enhance diagnostic accuracy. Leveraging the MIAS database, the m2D-CNN outperformed conventional CNN and feature extraction methods by effectively distinguishing between benign and malignant tissues. The proposed model achieved an accuracy of 95.29%, sensitivity of 0.93202, specificity of 0.9762, and an area under the ROC curve of 0.9718. These results demonstrate the efficacy of the m2D-CNN in early breast cancer detection, providing a scalable, robust solution for clinical application. This work underscores the potential of m2D-CNN in supporting radiologists in making informed diagnostic decisions.