Enhancing Radar-Based Stroke Detection: A Novel Approach Integrating Huygens' Principle and Deep Learning Techniques

Microwave imaging is rapidly gaining prominence as a favorable alternative to X-ray and MRI for various medical applications. In this study, we focus on investigating the Huygens' principle (HP) imaging method for brain imaging to detect strokes. The method has already shown promise in clinical trials for breast cancer detection. Adapting HP for brain imaging presents distinct challenges due to the brain's complex structure, necessitating the integration of advanced techniques. Current advancements in artificial intelligence and deep learning, despite their reliance on extensive datasets, offer valuable enhancements over traditional signal-processing methods. Through finite difference time domain (FDTD) simulations, we are able to generate the necessary comprehensive datasets, enabling the effective application of deep learning. Techniques such as U-net are explored for their ability to refine HP images into detailed stroke representations. Our findings through simulations demonstrate that integrating the magnitude and phase of HP imaging with deep learning methods significantly enhances stroke detection and classification. Specifically, using the magnitude of HP images resulted in an 83% accuracy rate in stroke classification, whereas utilizing the phase of HP images achieved an accuracy rate of 97%. Moreover, we successfully tested our model using data collected from two patients who participated in our first round of clinical trials.