Novel Microwave-Based NonContact Oedema Severity Monitoring with Machine Learning

Monitoring and detection of oedema severity is crucial for effective treatment of the underlying ailments. However, the conventional modalities for oedema detection either suffer from subjective assessment errors or involve setups that require harmful radiation, are expensive, tedious, and bulky. This work presents a novel, unobtrusive microwave-based sensing methodology utilizing the network analyzer, a customized antenna, and associated RF components for oedema severity monitoring. The RF patch antenna in the present situation is customized for operation at 4 GHz under dielectric loading of the numerical human phantom. The performance of the customized antenna is compared with that of the corresponding standard patch antenna at 4 GHz to emphasize its significance for oedema detection. The proposed scheme is validated by establishing an experimental setup emulating realistic oedema scenarios, and a diverse dataset is built by measuring the reflection data at the designated frequency under various conditions. For oedema severity classification, a machine learning (ML)-based approach is implemented making use of the extensive reflection data, and accordingly the physics-based features are extracted. Subsequently, by making use of the same, the ML assisted oedema severity (low, medium, and high) classification is carried out with a remarkable accuracy of 93.78 ± 4.15%.