ParKInsonPredict – Identification and prediction of disease progression patterns in Parkinson's disease

BMBF-project / Project start /

ParKInsonPredict is a joint research project of Technische Universität Dresden and Fraunhofer SCAI aiming to predict disease progression patterns in Parkinson's disease.

The increasing availability of digital biomarkers and large amounts of structured multi-modal data has the potential to fundamentally change the treatment and management of symptoms in Parkinson's disease. This includes the possibility of continuously recording symptoms in the home environment through sensors, digital questionnaires and interactive tests. However, this also involves structured digitally available data from clinical routine and clinical trials. These large and multimodal datasets offer the potential to uncover hidden patterns in the underlying data and thereby gain important additional clinical information that cannot be accessed using conventional approaches. One example is the prediction of disease progression patterns. Knowing the expected kind of disease progression encompasses not only important information about the individual patient but could also enable improved precise and individual treatment. Furthermore, such predictions would also future clinical trials to focus on fast progressors, i.e. patients with highest unmet need.

The application of AI algorithms in medical context is restricted by specific limitations of medical data sets. Also, most research focuses on data sets of single studies without validation on additional patient cohorts. Thus, the generalizability of the algorithms remains unclear.

The aim of ParKInsonPredict is to develop an AI model for the identification and prediction of disease progression patterns in people with Parkinson's disease. Based on preliminary work of Fraunhofer SCAI for multi-modal longitudinal data, an AI model will be developed and evaluated. Subsequently, the model will be adapted to a set of different Parkinson's disease cohorts, thus evaluating the generalizability of the AI model.

Project duration: 09/2022 – 02/2024