Enhancing Personalized and Transparent Health Metrics through AI

The increasing use of wearables like Garmin and Whoop has popularized the concept of 'body battery' scores, offering users a daily snapshot of their physical and mental readiness. However, these metrics are often generated within a black box, leaving the predictive validity and the underlying algorithms unclear to the user. This joint PhD proposal aims to address two key questions: (1) Do these black-box metrics have predictive value for physiological outcomes, such as performance during physical assessments or early markers of illness? (2) Can explainable AI (XAI) enhance these metrics, providing users with greater transparency and actionable insights? 

IMEC’s Speckle Plethysmography (SPG) technology is an advanced form of laser speckle contrast imaging that converts images into a 1D time series, similar to the current state-of-the-art photoplethysmography (PPG). SPG offers superior signal quality, particularly for individuals with darker skin and in varying ambient light conditions. Biomarkers derived from SPG include heart rate, heart rate variability, respiration rate, and potentially blood pressure and blood oxygen saturation. 

This research will explore whether integrating AI on top of these SPG-derived biomarkers, can produce a more accurate, personalized, and interpretable model of user health. A comparative analysis will investigate the predictive performance of commercial metrics (such as Garmin's and Whoop's body battery) and AI-enhanced models using SPG data. A key focus will be on understanding the physiological signals that drive these predictions and how they correlate with real-world performance and recovery metrics. 

From an IMEC perspective, the project will also assess the potential added value of in-house sensor technology. The goal is to validate whether IMEC's SPG sensors can outperform existing consumer-grade wearables in terms of accuracy, reliability, and providing users with meaningful health insights. If successful, this research could yield more transparent, data-driven health monitoring tools that will give users a deeper understanding of their physiological state.