Sensors in Medicine Textile Chest Strap Captures 110 Cardiovascular Parameters in Real-Time

Related Vendors

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

SIMTOS (Seoul International Machine Tool Show)

A research team from Fraunhofer IZM, together with partners, has developed a sensor system in the form of a textile vest designed to detect cardiovascular diseases early through continuous monitoring and AI analysis. It can simultaneously record over 110 medical parameters with a sampling rate of 1,000 data points per second each.

Cardiovascular diseases remain among the most common causes of death worldwide. This makes seamless and early monitoring of at-risk patients all the more important. As part of the Fraunhofer flagship project named "maia" (Medical Artificial Intelligence Applications), the Fraunhofer Institute for Reliability and Microintegration IZM in Berlin (Germany, together with Charité and TU Berlin, has developed a non-invasive sensor system that aims to elevate this approach to a new technical level.

Textile Chest Strap and Patch Sensors

At the center of the development are two flexible, body-worn wearable variants that wirelessly, synchronously, and at any time capture physiological data:

• A custom-fit textile vest with multi-channel electrodes that patients can quickly put on daily.
• A set of skin-friendly patch sensors placed on strategic body areas like the chest, neck, or legs to continuously collect data in the background.

This sensor combination enables the derivation of around 240 cardiologically relevant parameters. The system captures over 110 parameters simultaneously, with each individual measurement channel operating at a sampling rate of 1,000 data points per second. To create a holistic picture, the measurement results from external devices (e.g., ultrasound machines) can also be seamlessly integrated. According to the Fraunhofer IZM, this is the world's first system capable of capturing such a massive and synchronous volume of diagnostic data in everyday life.

From Classical Algorithms to Morphology Analysis

Given the enormous data streams, classical, manual evaluation methods fail in everyday practice. To maximize reliability, the Fraunhofer team relies on a multi-stage validation chain:

• The raw data from the sensors initially undergo classical signal processing methods.
• Subsequently, specialized AI models and neural networks are applied, trained to analyze the morphology of data curves, such as the course of ECG lines, in detail.

While such curve analysis in clinical practice usually requires meticulous manual effort, studies have shown that specialized AIs can perform this step even more accurately than experienced medical professionals. The AI thus detects subtle, dangerous trends indicating a deterioration in cardiovascular function.

Language Models (LLMs) as Clinical Co-Pilot

The technical concept, however, goes far beyond mere sensor data analysis: even anamnesis, such as family medical history or current daily condition, can be digitally integrated.

Precisely tuned Large Language Models (LLMs) with a medical focus are used for this purpose. Patients can answer guideline-based questionnaires via text and voice input; the survey is repeated daily for long-term observation.

In the background, a complex LLM ensemble operates, including Google's MedGemini. A superior AI agent cross-references the results of these medical language models and checks them for inconsistencies. Finally, the system tailors the results appropriately: depending on the target audience (medical professionals or patients), abnormalities, risk factors, and diagnostic suggestions are presented in understandable language.