Artificial intelligence for diagnostic support

Pilot project for the development of an intelligent algorithm to assist in the early diagnosis of psoriatic arthritis and axial spondyloarthritis

© | metamorworks

Psoriatic arthritis (PsA) and axial spondyloarthritis (axSpA) are immune diseases with a chronic course and often affect several organ systems simultaneously. In Germany, approx. 0.2 % of the population suffers from PsA (140,000 patients), axSpA affects approx. 550,000 people (0.8 % of the population).

Late diagnosis due to initially unspecific symptoms

Patients suffering from PsA or axSpA often complain of unspecific symptoms that may also resemble other conditions (e.g. painful joints of the hands/feet or back pain).  Many patients therefore only see a rheumatologist at a later stage of the disease, which worsens the prognosis. Due to the heterogeneous clinical pictures, a successful diagnosis and therapy depend to a large extent on the interpretation of the findings by the rheumatologist.

Potential of artificial intelligence

Various data sets from clinical routine are available for the medical interpretation of findings. Methods of Artificial Intelligence (AI) and automated image evaluation represent a great potential here. In addition, digitalization in the health care sector is a prioritized goal, which is, however, challenging in the conversion of unstructured data into information readable by algorithms.

Structured data sets through the use of artificial intelligence

The aim of this project is to convert predefined, typical clinical data sets into structured data sets by using AI and to evaluate collected image data (X-ray, MRI) automatically. The evaluation of findings based on frequencies will then be used for medical interpretation to generate an output that summarizes and presents a first prediction of the disease symptoms for review.

The first cornerstone for the implementation of such an algorithm in the care system will be established in this project. In the long term, it should be possible to evaluate in a supportive manner at the level of the family doctor whether the patient presented has an increased probability of the presence of PsA or axSpA with regard to the constellation of findings and symptoms and should therefore be sent to a rheumatologist for clarification at an early stage.

Interdisciplinary cooperation

Fraunhofer ITMP:

Medical interpretation of data sets, conduct of clinical studies, expertise in the field of immune-mediated and inflammatory diseases

Fraunhofer IAIS:

System development based on artificial intelligence and application to medical data

Fraunhofer IGD:

Development of software and algorithms for the representation and automated evaluation of different imaging techniques

Fraunhofer IMW:

User centricity


With the successful completion of this project, a first demonstrator will be developed, which creates pipelines for the collection of clinical data in the indications PsA and axSpA as well as for their conversion and evaluation into structured data sets. This demonstrator will then be further developed and specified for use in a larger study project in clinical routine. The aim is to develop a supporting algorithm that increases the rate of correct referrals to the specialist by supporting it with AI. By contacting health insurance companies, the implementation of the algorithm in existing digital health care systems can be discussed in the course of the project in order to offer the algorithm throughout Germany. First contacts with representatives of health insurance companies have already been established and the interest in pursuing the project has been expressed.