A machine learning assisted eight-element metamaterial antenna array for non-invasive breast cancer detection

Abstract

Breast cancer remains one of the most prevalent cancers and a leading cause of cancer-related mortality among women worldwide, highlighting the need for accurate and non-invasive early detection methods. This study presents an integrated breast cancer diagnosis system that combines a microwave imaging platform composed of eight metamaterial-based antenna elements with machine learning (ML) techniques for breast tissue classification. The proposed non-ionizing approach utilizes backscattered S-parameters collected from the antenna array over the 3-7 GHz frequency range to identify tumor masses with diameters of 20 mm, 25 mm, and 30 mm. A dataset comprising 32,048 samples was generated from the measured S-parameter responses and categorized into four classes: healthy breast tissue and tumor cases with diameters of 20 mm, 25 mm, and 30 mm. Three ML classifiers, namely adaptive boosting (AdaBoost), gradient boosting (GB), and random forest (RF), were trained and evaluated using metrics including classification accuracy (CA), Matthews correlation coefficient (MCC), F1-score, recall, and precision. Experimental results demonstrate the effectiveness of the proposed framework, with AdaBoost achieving the best performance and an accuracy of 99.8%, outperforming the other classifiers. The findings confirm the potential of combining metamaterial-based microwave imaging with ML techniques for accurate, non-invasive, and early breast cancer diagnosis.