AI-Integrated Noninvasive Microwave Sensor System for ARDS Diagnosis

Acute respiratory distress syndrome (ARDS) is a lung condition that poses a high risk of respiratory failure. As conventional diagnosis tools, CT scans and X-ray scans lack portability, low cost, and safety due to their ionizing nature. This article presents a portable, low-cost, safe AI-integrated microwave sensor system designed for the diagnosis and severity assessment of ARDS. The system incorporates an AI processing unit equipped with an extreme gradient boosting (XGBoost) classifier to analyze and classify the scattering parameters obtained from sensors into four distinct severity categories. This classification is crucial for accurately assessing the condition based on the sensor data. In the validation process, a liquid phantom mimicking the chest region, along with porcine respiratory organs, is used to fine-tune the trained classifier using incremental learning methods and to verify its performance. Our system presents a promising tool for diagnosing and continuously monitoring ARDS severity, showing high accuracy in tests. The development of this portable microwave sensor system addresses a critical need for more efficient and precise severity assessment in ARDS patients, offering potential contributions to biomedical research in early diagnosis and respiratory care.