AI-powered heart monitoring

Smart Wearables Instead of Cable Spaghetti

More than nine million people worldwide die each year as a result of heart disease. In Germany, it is responsible for a third of all deaths. An AI-supported sensor system can detect warning signs at an early stage, saving lives.

Exhaustion after jogging, shortness of breath when climbing stairs, anxiety before falling asleep: Most of the first signs of heart failure are nonspecific. Those affected therefore often put off seeking medical help for a long time. “And when they finally do go to the cardiologist, maybe everything looks just fine. But the next day, when they go for a run, the problems come back because the heart doesn’t reach its limit until it’s under stress,” says Basel Adams, team leader at the Fraunhofer Institute for Reliability and Microintegration IZM. Comprehensive heart monitoring is often not available until after a wait of several weeks and involves tremendous effort.

Detecting heart failure early

“There must be a better way,” thought Adams, a trained electrical engineer and got to work on his dissertation. In the Fraunhofer maia (Medical Artificial Intelligence Applications – Center for Applied AI in Medicine) joint research project at Fraunhofer IZM, he worked together with partners from Charité Universitätsmedizin and TU Berlin to develop an intelligent sensor system that can be worn directly on the body. It measures a large swath of cardiac and circulatory data – wirelessly, in real time and at any time – and uses this data to derive AI-supported diagnoses and treatment recommendations. To date, two types of sensor-equipped wearables have been developed. One is an individually adjustable textile vest designed to be worn once a day for 10 to 15 minutes over an extended period and uses built-in sensors to record physiological data. The other is an ensemble of biocompatible patches that are placed on various parts of the body, over the heart and lungs, on the neck and on the legs to continuously gather data.

Textile vests and sensor patches: Flexible wearables for everyday use

Data acquisition in the wearables is performed by multi-channel semi-dry textile electrodes that can distinguish relevant signals from interfering body movements. The biocompatible electrodes are integrated in the washable fabric vest with hook-and-loop fasteners for optimal contact on the body depending on the wearer’s size and build, contact the skin directly without any conductive gel.

Whether in the form of a vest or with patches, the smart sensor system measures a wide swath of medically relevant signals, including body temperature, respiratory rate, heart and lung sounds blood pressure and oxygen saturation, stroke volume and cardiac output, a complete 12-channel ECG and much more. The total of 40 acquired raw data signals can be used to derive 110 cardiological parameters. They provide a comprehensive picture of cardiovascular function during various activities, whether the person is awake or asleep, exercising or experiencing stress, feeling happy or upset. To integrate this impressive arsenal of state-of-the-art medical technology into comfortable, everyday wearables, the various instruments first had to be miniaturized and connected together through interfaces. The result is a world first, stresses Basel Adams: “Although there are all sorts of sophisticated devices on the market, they are far too large and generally don’t provide any raw data but rather only derived results. Nor do they have any standardized interface because every manufacturer uses its own system.”

The new wearables can synchronously record and compare all this data at a sampling rate of 1,000 points per second. But that’s not all. “The system also receives input from two additional data sources,” explains Adams. A cardiologist first creates a patient profile with all the relevant information about known hereditary conditions, previous surgical procedures and medication used. “We also ask the person how they feel at present, whether they tire easily or if they have noticed any chest pain, difficulty breathing or other irregular symptoms.” A chatbot obtains this information by asking questions for five to ten minutes once a day, as would a doctor, and saves the answers along with the other data in the cloud. 


Comprehensive heart monitoring with 1,000 data points per second

Each measurement thus combines more than a thousand data points from three sources, serving as the basis for a diagnosis or risk analysis by the system. “The key is that we record all of these signals synchronously. The AI can thus learn to recognize patterns and draw the right conclusions from them. This is the diagnostic strength,” explains Adams. He first obtained the necessary expertise in in-depth dialogue with Charité cardiologists and then used this expertise to train the AI. The guidelines of the German Cardiac Society serve as the basis. “We want to ensure in this way that our system follows the same criteria as a doctor would in practice.”

Before the sensor-equipped vests and patches can be tested on people with heart disease, their effectiveness has to be demonstrated in clinical trials. The question: Is the AI-supported system correct in its diagnoses and treatment recommendations? Or does it even make more accurate assessments than experienced cardiologists? While Basel Adams looks for partners for studies – the necessary approval from the ethics committee has since been provided – his colleague Lukas Werft is already considering further improvements to the wearables. In the Fraunhofer IZM System On Flex group, the materials engineer played a key role in developing textiles with integrated flexible electronics. In the current model of the vest, the various electrodes transmit their signals via conventional cables that are routed between two layers of fabric. “Once our product is launched on the market, we could consider integrating the electrodes in a T-shirt, for example,” he says. “It should be possible to just toss it in the washing machine after wearing it, ideally as often as needed. And of course it would be practical to not have to first remove all the cables every time.”
 

Liquid Metal Ink for smart textiles

The researcher already has an idea of how to solve this problem: with a printable and washable liquid-metal ink that he developed together with colleagues at Fraunhofer IZM. The main component of this innovative Liquid Metal Ink (LMI) is Galinstan, an electrically conductive alloy of gallium, indium and tin. It can be printed on and encapsulated in thermoplastic polyurethane, a plastic that can be shaped using heat, and then laminated on standard textiles. “If we could replace the conventional leads with this liquid metal, sensor-equipped textiles such as T-shirts or vests could be extremely durable and washable,” underscores Werft.

Lukas Werft sees another medical application for this liquid metal ink in pediatrics: “Unfortunately, far too many babies still die from sudden infant death syndrome because they stop breathing unnoticed during the night. Traditional monitoring systems are uncomfortable and awkward. Our approach would be to integrate Galinstan-based strain sensors directly into baby clothes.” A prototype of a sensor-equipped garment for babies has already proven its function in a Europe-wide research project called Newlife.                    

 

Science Year 2026 – Medicine of the Future

Fraunhofer Group for Health

The goal of the Fraunhofer Group for Health is to foster a high degree of interdisciplinarity across the 4D model – “Drugs,” “Diagnostics,” “Devices,” and “Data.” The potential for innovation resulting from the integration of various scientific disciplines is intended to generate tangible added value for health research and patients.