Non-invasive HealthCare

In the aforementioned applications, numerous sensor data are usually collected, processed and evaluated. This can generate extensive amounts of data for which there is insufficient transmission bandwidth to a cloud application or for which a long latency period until evaluation is not permissible. In such cases, there is the possibility of using algorithms that insist on artificial intelligence (Embedded KI), e.g. with neural networks, to implement complex data processing and evaluation directly at the point of collection. This does not necessarily require complex hardware with high computing power; a small energy-efficient microcontroller is sufficient in many use cases. In some cases, special AI hardware accelerators are needed so that the algorithms can work even faster and more efficiently.

One solution is optical contactless vital parameter measurement. The Fraunhofer IMS is researching within the application field of Non-invasive Healthcare the most accurate and fastest possible detection of these vital parameters in different wavelength ranges and quantifies them using machine learning methods. Our novel approach to non-contact measurement uses camera-based photoplethysmography (PPG). The physical basis of this technique is the different absorption or reflection spectra of human skin. By measuring and analyzing the time course of color, spectral and temperature values of the human skin, various vital parameters can be determined. For example, heart rate can be measured using color image analysis and extraction and filtering of PPG signals. Through in-depth analysis of RGB (red, green, blue) PPG signals, respiratory rate and oxygen saturation can also be determined without contact.

Furthermore, the Fraunhofer IMS is working within the application field of Non-invasive Healthcare on the next generation of sensors close to the body, so-called smart patches , which can be worn comfortably. The analysis is performed directly on the patch by means of integrated artificial intelligence. Local processing with sufficient detection quality is not yet possible with conventional signal processing, since digital filters in particular require a high computing capacity. This is accompanied by a large size, high weight, high energy consumption, low battery life and increased costs. A solution is provided by local artificial intelligence (Embedded AI) which performs analysis directly on the patch and is resource-efficient. For integration into battery-powered devices in long-term use, processing must be handled on resource-limited hardware. Special AI hardware accelerators are needed to make the algorithms work even faster and more efficiently.

When brainstorming, designing and developing sensor assistance systems, the Fraunhofer IMS has the option of using components that are already available as well as developing new components based on IMS technologies and integrating them into a complex system to offer solutions from a single source. These IMS technologies include special sensors, customized electronic components, radio systems und Embedded Systems (hardware and software).

Together with the customers we bring - if required - the develop up to CE-certification. A successful example of an approved medical device is the electronic assistance system EQUIVert which was developed especially for patients with dizziness symptoms, in order to support the physician in the dizziness diagnosis and therapy finding and to relieve him in the dizziness therapy of the patients, as well as to enable the patient an independent regular training.

In the Living Lab of the Fraunhofer-inHaus-Center with its hospital engineering and intelligent care room, the novel electronic assistance systems can be evaluated under realistic conditions, optimized and thus best tailored to the application and the market.