In this project, prof. dr. Maurits Kaptein and dr. Marta Regis from the Eindhoven University of Technology, will collaborate with Omron Healthcare Digital Health Labs. Their goal is to develop MCP-Health, a Model Context Protocol (MCP). This will be an open-source standard for connecting AI applications to external systems that transforms streams of longitudinal physiological measurements (like heart rate, blood pressure, sugar levels) into compact, information-dense summaries which can be processed more effectively by an LLM.
Large language models (LLMs) are increasingly being explored for interpreting longitudinal health data, for example, continuous blood pressure readings from wearable devices. However, directly providing raw longitudinal or time-series data to an LLM is generally both unreliable, due to lack of robust evaluation of patient trajectories and poor uncertainty estimation, and computationally expensive. Long sequences consume large numbers of tokens without guaranteeing improved clinical reasoning, and noisy observations further degrade output quality. A principled mathematical approach to compressing health time-series into token-efficient representations is therefore both timely and necessary.
This project directly demonstrates the foundational role of applied mathematics in making AI systems reliable and efficient for healthcare. The core challenge is how to compress highdimensional, noisy time-series into informative, uncertainty-aware representations. This challenge is genuinely mathematical, involving statistical signal processing, information theory, and (Bayesian) uncertainty quantification.
The collaboration with Omron Healthcare ensures that the mathematics is driven by and validated against real industrial and societal needs. The resulting open-source tool will make the mathematical contributions visible and reusable beyond academia, and the project creates a concrete valorization pathway from TU/e's mathematics research to digital health applications. The involvement of a 4TU mathematics institute (TU/e) in co-creating an industry-ready prototype exemplifies the mission of 4TU+.AMI.
Contact persons
- Prof. dr. Maurits Kaptein (TU/e)
- Dr. Marta Regis (TU/e)