Overview

Sensor data for predicting health status and supporting medical decisions by continuously recording relevant vital parameters are becoming increasingly important. The dynamics of complex diseases and the effectiveness of therapeutic and medical technology interventions can often only be adequately tracked through the use of continuous monitoring combining various vital parameters. In healthcare, the fusion of sensory vital signs on a sensor platform holds particular potential for the use of artificial intelligence (AI) to support medical decision-making and determine early warning signals that would otherwise remain hidden by conventional diagnostic methods. The use of AI technologies opens up possibilities for prognosis and (semi-) autonomous diagnosis of health conditions, which may, however, radically change healthcare and the traditional doctor-patient relationship.

The foundation of AI-based prognosis is the precise and continuous acquisition of data from multiple sensors. One measurement point on the body that is ideally suited for this is the ear. Here, modern hearing systems can be extended to include non-invasive sensor systems. These include temperature sensors, inertial sensors for measuring acceleration and movement, optical sensors for measuring oxygen saturation, heart rate and heart rate variability, and EEG sensors for measuring brain activity. The ear provides an accurate location suitable for long-term measurements, where a sensor-equipped hearing system can be integrated unobtrusively into patients’ daily lives. Through a human-machine interface, the medical staff and the patient are able to interact with the system and the AI. Acoustic interaction and a utilization of the basic functionality of a hearing aid are of particular importance.

The close-meshed collection and evaluation of sensor data and their use for data-based decision support systems are the basis of a wide range of medical technology and social innovations. Decision-support systems in healthcare are thereby able to deal with the exponentially increasing data combinations captured by sensors. The linking of sensor data collected on and in the ear with artificial intelligence approaches is of great importance for the optimization of care processes in various medical application fields:

• Audiology: The sensory determination of the individual stress level (hearing effort) and hearing fatigue using AI methods enables the development and marketing of applications by audiologists and hearing aid manufacturers for automatic hearing aid adjustment, which reduce the hearing effort of the hearing aid user in everyday life and thereby avoid hearing fatigue.

• Occupational medicine: A potential application area beyond audiology is in the context of occupational medicine, e.g. to minimize health risks and for occupational safety in safety-critical task areas. In the presence of health risk factors, for example, the state of fatigue, stress levels or cardiovascular stress can be determined and used in occupational medicine to implement preventive occupational safety measures and avoid overloads by recognizing and communicating critical situations.

• Monitoring of high-risk patients: Continuous monitoring of high-risk patients, especially in the field of cardiology to detect cardiac abnormalities, supports medical staff in diagnosis, treatment planning and identification of risks. A smart sensor system can also process information directly for patients through appropriately designed user interfaces, thereby supporting patients’ ability to act independently (patient empowerment).

These three fields of application benefit from continuously increasing data volumes and the training of complex algorithms on highly parallelizable hardware architectures. Since a ready-trained AI algorithm does not require large hardware resources, it is possible to run them on comparatively low-power hardware systems, such as hearing aids or cell phones, in real time and in an energy-efficient manner. Through these technological developments, it is also possible to use AI algorithms locally and mobile on humans to improve hearing quality, support secondary prevention in occupational medicine, and establish continuous health monitoring with integrated interaction.

The potential of AI-based sensor systems faces challenges, particularly in the healthcare sector, with regard to changes in traditional care and work processes as well as risks in the highly regulated healthcare market. Therefore, the technical development and the design of reference models for human-machine interaction will be aligned with social, ethical and legal requirements. In the project, the preferences of medical staff and patients will be determined on the basis of qualitative and quantitative empirical studies and supplemented by a normative analysis of the ethical and legal framework with regard to the various roles, competencies and responsibilities in the socio-technical system. By aligning system development with social and ethical requirements, increasing the decision transparency of AI, and using appropriate human-machine interactions, acceptance on the part of patients and medical staff is ensured in the early stages of development.