![](\"https:\/\/cdn.pblzr.de\/dacbb27c-1270-4041-b681-e2b95f06f8a1\/2026\/08\/fraunhoferizm-logo-400x300_TEXT.jpg\"){kind=logo}
<\/p>\n<\/p>\n
Even for the most experienced doctors, it is nearly impossible to keep track of everything when it comes to diagnosis: According to the WHO, there are 55,000 distinct diseases with 13,000 symptoms.1 Among these, cardiovascular diseases\u2014such as high blood pressure, heart failure, and arrhythmias\u2014are among the most common. They account for a large proportion of deaths worldwide.<\/p>\n
But they are well-researched and recognizable. Wearables, such as smartwatches, often contain sensors that alert their wearers to warning signs. They do not replace comprehensive cardiological diagnostics. However, it often takes a long time to get an appointment with a specialist, and in the doctor\u2019s office itself, there is little time for individual patients.<\/p>\n
Under these circumstances, artificial intelligence (AI) offers a way to obtain a complete picture of a patient\u2019s condition and a reliable diagnosis. However, efforts to establish an AI framework for cardiological diagnostics have so far encountered a chicken-and-egg problem: Since no device existed that could guarantee the measurement of all relevant parameters, there are no AI systems specifically trained on cardiological parameters. Without specialized AI, however, a complete diagnostic system is not possible.<\/p>\n
Thanks to a sensor system developed at the Fraunhofer Institute for Reliability and Microintegration IZM, a solution to this problem is now within reach. The data collected by the sensor system is transmitted to the cloud via an edge PC integrated into the wearable device and evaluated there with the support of AI.<\/p>\n
Approximately 240 cardiologically relevant parameters can be derived from the measurements of the built-in sensors. In addition, measurement results from external devices, such as ultrasound machines, can be fed into the system. To optimize reliability, the data is first validated using traditional signal processing methods and then using AI models trained on this data.<\/p>\n
Neural networks can then analyze the morphology of the data curves\u2014such as the shape of the ECG waveform\u2014to identify abnormalities. This analysis is still performed manually and represents a necessary but time-consuming task in clinical practice. However, studies have repeatedly shown that specialized AI systems can perform this analysis more accurately than experienced medical professionals.<\/p>\n
Added to this are insights from the patient\u2019s medical history\u2014such as family medical history or their current daily well-being. Through the use of precisely tailored Large Language Models (LLMs)\u2014language models focused on medical topics\u2014these assessments can be conducted in detail and repeated daily for long-term monitoring. The questionnaire, which is based on cardiology guidelines, can be completed via text or voice input. LLMs are also used in the processing of examination results. Tailored to the specific user\u2014whether medical professionals or patients\u2014anomalies, risk factors, and diagnostic suggestions are presented in language that is easy to understand. To this end, several medical LLMs operate in the background\u2014including Google\u2019s MedGemini\u2014whose results are cross-checked by a higher-level AI agent and reviewed for inconsistencies.<\/p>\n
By supporting standard processes in this way, specialist care could be organized more efficiently in the future. Specialized systems would assist primary care physicians in the early detection of symptoms and risks and facilitate patient-centered referrals to the appropriate specialists. Specialists could draw on a far more comprehensive dataset and receive support in its evaluation and interpretation. The system can support patients living with cardiovascular disease by adjusting the correct medication dose based on their daily condition.<\/p>\n
By supporting standard processes in this way, specialist care could be organized more efficiently in the future. Specialized systems would assist primary care physicians in the early detection of symptoms and risks and facilitate patient-centered referrals to the appropriate care provider. Specialists could draw on a much more comprehensive dataset and receive support in its evaluation and interpretation. The system can support patients living with cardiovascular disease by adjusting medication doses to match their daily condition.<\/p>\n
The further development of the sensor system toward AI-supported validation, evaluation, and processing of measurements offers a glimpse into the \u201cmedicine of the future,\u201d in line with the motto of the Year of Science 2026. In the future, the use of similar systems is also conceivable beyond cardiology, for example in pulmonary medicine.<\/p>\n
(Text: Steffen Schindler)<\/i><\/p>\n
– – – – –<\/p>\n