Academia Sinica Logo

Biomarkers have traditionally been defined by biochemical or physiological measurements obtained through direct sampling or physical contact. However, recent advances in engineering and data science are fundamentally transforming this concept. This lecture introduces a new perspective on biomarkers through the lens of engineering-driven digital medicine. I will present a series of research efforts that demonstrate how multimodal biosignals—such as electroencephalography (EEG), camera-based photoplethysmography, and autonomic nervous system signals—can be translated into clinically meaningful digital biomarkers without physical contact. These approaches enable real-time estimation of physiological and mental states, including cardiovascular dynamics, sleep stages, stress, and emotional states, under real-world conditions. The lecture will highlight key methodological innovations in signal processing and machine learning that ensure robustness, interpretability, and clinical relevance. In addition, translational aspects will be discussed, including validation strategies, regulatory considerations, and pathways toward clinical and societal implementation. By bridging engineering innovation with medical needs, this talk aims to redefine the role of biomarkers in digital medicine and to explore how non-invasive, scalable technologies can support future preventive and personalized healthcare.