Data-driven analysis of medical and simulation data for improved patient treatment in rhinology

In a joint cooperation with Jülich Supercomputing Centre (JSC), Forschungszentrum Jülich, and Park-Klinik Berlin Weißensee, new methods for the analysis of patient and simulation data will be developed. These methods will support decision processes a priory a rhinological surgery and will allow to predict surgery outcomes. Therefore, novel numerical methods will be developed to enhance a knowledge base consisting of medical data by simulative fluid-mechanical data. The data base will be structured and enhanced by intelligent labels to be optimally usable by machine learning (ML) algorithms for training. New ML techniques in the field of Deep Learning (DL) such as Tensorflow/Keras and Convolutional Neural Networks (CNNs) will be evaluated. These techniques will be employed to extract hidden data from medical and simulative data, to classify pathologies, and to support surgery decisions. Furthermore, the success probability of a surgery will be derived based on historical data. This hybrid approach will be the first of its kind to couple real patient data with simulative data, and ML predictors in a full loop that reintegrates new information into the process chain. The project collaborates with the ZIM-funded project Rhinodiagnost ( in which the corresponding numerical methods for the simulation of respiratory flows are developed. An intense exchange with the experts from Park-Klinik Berlin Weißensee guarantees an adaption of the methods to the medical requirements.