In the project SATURN, which is funded by the German Federal Ministry of Health, Fraunhofer IESE is collaborating with its project partners to research methods for using AI technologies from the areas of rule-based systems and machine learning for traceable and transparent decision support / tentative diagnosis in the context of unclear diseases.
Last modified:
The development and establishment of a Smart Physicians’ Portal for Patients with Unclear Disease (SATURN) supports the diagnosis of patients based on Artificial Intelligence (AI) and provides a traceable and transparent tentative diagnosis. Different AI approaches based on expert knowledge or clinical case data are used. The diagnosis support based on rule-based approaches is based on medical expertise extracted by experts from available knowledge sources, such as guidelines and specialist literature. AI experts at Fraunhofer IESE then develop rule-based models based on this knowledge. In contrast, diagnosis support by means of machine learning and case-based reasoning uses real clinical data from university hospitals, taking into account data protection and patient privacy. The challenge here is that the amount of data is sometimes small. For this reason, hybrid approaches are also being investigated, where models are developed and refined both on the basis of expertise and on the basis of data. In all developed methods, special attention is paid to transparency and traceability for the users. Therefore, approaches from the areas of Explainable AI and uncertainty assessment are being used. The goal is to strengthen users’ trust in the AI-based solution. If the tentative diagnosis indicates a rare disease, SATURN enables referral to experts from university hospitals. With the help of guideline-oriented standards of care, specific recommendations are given for common diseases.
Fraunhofer Institute for Experimental Software Engineering IESE