AI-Assisted Sensor Data Analysis for Cardiac Diagnostics

A sensor system developed at the Fraunhofer Institute for Reliability and Microintegration IZM aims to support the diagnosis of cardiovascular diseases. The approach combines wearable measurements with AI-assisted data validation and analysis, alongside information from the patient’s medical history. It is intended to provide healthcare professionals with a broader dataset and support interpretation.

Wearable Sensors Provide Around 240 Cardiac Parameters

Approximately 240 cardiologically relevant parameters can be derived from the integrated sensors’ measurements. Results from external devices, such as ultrasound systems, can also be incorporated. An edge PC integrated into the wearable transfers the collected data to the cloud for AI-assisted analysis.

To improve data reliability, conventional signal processing is followed by validation using specially trained AI models. Neural networks can then examine the shape of measurement curves, such as ECG traces, to identify abnormalities. The latest system version uses reusable sensor patches and a wearable computer.

Medical Language Models Support Patient History and Findings

Medically focused large language models (LLMs) help collect additional information, including family medical history and the patient’s current condition. A questionnaire based on cardiological guidelines can be completed by text or speech and repeated daily for longitudinal monitoring.

LLMs also help present the results. Findings, risk factors and diagnostic suggestions are explained in language tailored to healthcare professionals or patients. Several medical language models contribute to the process, while a supervisory AI agent compares their outputs and checks for contradictions.

In the longer term, the approach could support early detection in primary care and help specialists analyze and interpret larger datasets. Applications beyond cardiology, such as respiratory medicine, are also envisaged.

Fraunhofer IZM will exhibit with IVAM at COMPAMED 2026 in Düsseldorf, November 16–19.