Projects

Early detection of respiratory diseases using infrared thermography and AI methods

Common symptoms of Covid-19 disease are fever and respiratory impairment, which are expressed by coughing or changes in breathing patterns (e.g., accelerated, shallow breathing). Both can be detected remotely using non-contact sensor technology. Infrared thermography is of particular importance here: it is independent of illumination and the visual impression of thermography is very different from video recordings in the visible range. This facilitates anonymization and improves user acceptance, which enables the application in public spaces. The project will develop artificial intelligence methods that enable early detection of Covid-19 disease from infrared thermography data.

A federated paradigm of real-world data sources utilization for the empowerment of diagnosis, prognosis and risk assessment of cardiovascular conditions

CVDLINK brings together 19 partners from 10 European countries and Israel with a budget of 9.7 million euros. Project partners include universities, research institutes, and companies from Lithuania, Germany, Ireland, Greece, Austria, Finland, Spain, Romania, Moldova, and Estonia. The four-year project is funded by the EU Research and Innovation program 'Horizon Europe', and KIS*MED leads the work package 'Explainable, robust and trustworthy AI tools' within it.

Every year, more than 6 million new cases of cardiovascular diseases (CVDs) are reported in the European Union, causing 1.8 million deaths. This poses a significant burden on healthcare systems and society. The CVDLINK project aims to address general challenges related to patient health data by developing a Europe-wide software platform service for effective data-driven prediction and prevention measures. The platform, based on federated learning principles, will integrate existing and future sources of medical data while ensuring data security and privacy.

Biomechanically informed, trustworthy machine learning from social video platforms to monitor physical training

Social video platforms such as YouTube are a popular source for learning various skills, including physical exercise. We want to develop a system that makes use of this abundant source of information to help users perform exercises correctly. In the bigger picture, we want the machine to automatically learn what is correct exercise from the “wisdom of the crowds” contained in social video platforms. For this, we propose a biomechanicsinformed machine learning approach. By fusing computer vision, biomechanical modeling and machine learning, the system will be rooted in meaningful biomechanical parameters and hence be efficient in terms of computational demand and data required. Most importantly, it will be intrinsically explainable and thus will be able to classify exercise quality and give recommendations in a trustworthy way. In this interdisciplinary project, KIS*MED will contribute with expertise in unobtrusive sensing of motion, signal processing, and machine learning, while Lauflabor locomotion lab will contribute with expertise in biomechanical modeling, sport science, and recording of human motion. The project is funded by FiF – Forum for Interdisciplinary Research.

Yurt Dışına Lisansüstü Öğrenim Görmek Üzere Gönderilecek Adayları Seçme ve Yerleştirme

YLSY (“Selection and Placement of Candidates Sent Abroad for Postgraduate Education”) is an official grant offered by the Republic of Türkiye Ministry of National Education to provide qualified postgraduate students for public offices and agencies.

At KIS*MED, Gökhan Güney is a YLYS scholarship holder to pursue his Ph.D. studies. His work focuses on artificial intelligence-assisted human state analysis in a broad spectrum, including, but not limited to, Parkinson’s patients’ hand movement and eye blink reflexes, and the emotional states of healthy individuals.

A multimodal simulation environment for camera-based monitoring of cardiorespiratory activity.

A popular theme in many future scenarios is the doctor who can scan a patient remotely and immediately assess their overall health status. In practice, camera-based methods have shown promising approaches to approaching this vision. Particularly for monitoring breathing and pulse, KIS*MED has already researched techniques to capture pulse rate and respiratory rate using cameras under optimal conditions. To advance further development, a wide variety of diverse datasets are needed.

However, the major challenge is that camera images represent protected patient information, are practically non-anonymizable, and therefore difficult to share among different research groups. Furthermore, for many medical conditions, if any, there are only very few measurement data available that are not sufficient for evaluating existing algorithms. Therefore, the DFG-funded project aims to create a simulation environment capable of generating synthetic video data of realistic avatars exhibiting relevant cardiorespiratory activity. With these data, novel algorithms can be tested, deep neural networks can be trained, and for the first time, they enable easy exchange among international research groups.

In this project, which is funded by the Hessian Ministry of Digital Affairs, we want to work with smart medication eHealth Solutions GmbH to develop methods and concepts for using AI and non-intrusive sensor technology to carry out quantitative and qualitative movement analysis for physiotherapy exercises for patients with haemophilia.

Haemophilia is an inherited rare chronic disease in which blood clotting is impaired. In Germany, around 4,000 patients are affected by severe haemophilia. In patients with haemophilia, spontaneous bleeding can occur all over the body, even without injury and despite therapy. Joints, especially ankle joints, knees and elbows, are particularly often affected and their deterioration can cause pain, limited mobility and, in the further course of the disease, joint destruction. Patients therefore often receive orthopedic and physiotherapeutic treatment. The aim is to maintain joint mobility.

Adherence to therapy is always a challenge in the treatment of chronic patients, especially adolescents in puberty. For some years now, digital solutions such as electronic diaries have been used for (tele)monitoring and therapy in the form of digital health applications with medical apps. In addition, fitness trackers and smartwatches are being used as general approaches for monitoring patient activity. The disadvantage of these approaches is that they do not allow differentiated recording of movement with qualified statements about movement sequences of individual joints.

Based on this situation, we will work with smart medication eHealth Solutions GmbH, an established industry expert, to develop methods and concepts that enable such qualitative and quantitative movement analyses using AI (machine learning) and non-intrusive sensor technology (camera, wearables). Furthermore, serious games concepts and gamification principles are to be developed with the Serious Games Group of TU Darmstadt (PD Dr. Stefan Göbel), which support targeted personalised (physical) therapy programmes and playfully contribute to increasing therapy adherence.