Machine Learning in MEMS for Biomarkers Generation

点击翻译按钮获取中文摘要

The goal of the proposed project is to integrate the new AIMEMS sensor technology with wearable medical device to improve the generation of digital health biomarkers. AIMEMS form an entirely new class of MEMS, where AI capabilities are built directly into the dynamics of the MEMS. This tight integration of AI into physical devices allows the fabrication of compact and energy efficient sensors, that can be taught using standard machine learning algorithms to realize complex data filters, classifiers or other computational tasks. The principal new concept in the proposed research is to use the AIMEMS machine learning capabilities 'at the edge' of a digital biomarker data processing system. These biomarkers are physiological and behavioral measures collected through connected digital tools to explain, influence or predict health-related outcomes. The outcomes of the proposed project will comprise new or improved digital biomarkers with a strong potential to generate medical breakthroughs for the diagnosis and management of viral respiratory diseases (such as COVID-19 and Influenza), and of type 2-Diabetes Mellitus with cardiovascular diseases as its principal comorbidity. The research activities will be organized along two mostly parallel tracks: one that uses an AIMEMS microphone to identify and classify different acoustic signatures from respiratory infections, and one that uses an AIMEMS accelerometer to improve the efficiency of a wearable digital biomarker system for the screening of at-risk individuals, to promote the diagnosis and management of type 2-Diabetes Mellitus and cardiovascular diseases.