SRR-Based Microwave Sensing for Lymphedema Detection: A Machine Learning Approach

In this research, we explore the potential of microwave sensors for early detection of Lymphedema (LE). Lymphedema is a chronic condition characterized by the accumulation of protein-rich fluid in the tissues due to impaired lymphatic drainage. The condition can result from genetic abnormalities of the lymphatic system, damage to the lymphatic system due to surgery, radiation therapy, infection, trauma or other causes. While various medical solutions have been investigated to prevent and detect this condition, often involving expensive tests such as Magnetic Resonance Imaging (MRI) or Computerized Tomography (CT), the application of microwave sensors offers a promising, efficient, and cost-effective alternative. Microwave sensor measurements, combined with Machine Learning techniques, may provide a viable solution for understanding the presence or stage of the disease, as well as the monitoring of its evolution. This thesis analyzes readings obtained using a Split Ring Resonator (SRR) on patients who have undergone surgery. The goal is to perform binary classification to determine whether the sensor can detect the presence or absence of lymphedema. This is achieved by analyzing resonance frequency shifts and penetration depth. Various machine learning algorithms are tested, and two types of classification are conducted: one between healthy limbs and reference limbs (belonging to patients with LE), another by assessing the difference between the two limbs for each individual. The latter approach accounts for individual characteristics that may influence the measurements, such as skin thickness and muscle mass. The results are promising, particularly when considering both the difference between the two limbs and the location of lymphedema, yielding sensitivity of 93% and accuracy of 90.35%