Multi-Tiered Self-Contrastive Learning for Medical Microwave Radiometry (MWR) Breast Cancer Detection

Breast cancer, marked by the uncontrolled and rapid growth of cells due to genetic mutations, significantly impacts global health, as it records one of the highest incidence rates of cancer. In 2020 alone, it was estimated to account for 2.3 million new cases, becoming the primary cause of death among women with nearly 700,000 deaths [1]. Disturbingly, future forecasts suggest a continued rise in both the occurrence and death rates associated with breast cancer [2]. The pivotal role of early detection in reducing mortality rates and reducing the healthcare load cannot be overstated. In this context, Microwave Radiometry (MWR) emerges as a promising imaging modality that passively captures the natural microwave emissions of human tissues [3]. Its utility spans a broad spectrum of clinical areas, including but not limited to, the breasts [3, 4, 5, 6], brain [7, 8], lungs [9], veins [10], and musculoskeletal structures [11]. Within the domain of breast cancer screening, MWR leverages the fact that cancerous tissues, due to their increased metabolic rate, emit more heat than normal tissue [4].