Using digital clinical and research data to develop a predictive model related to hearing outcome after cochlear implantation.

• Applicants: Dr. Benedikt Hoeing; Professorin Dr. Christin Seifert
• Subject Area: Otolaryngology, Phoniatrics and Audiology ; Medical Informatics and Medical Bioinformatics
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 536124560

Introduction: Cochlear implants (CI) enable deaf people to perceive sound. The amount of pre- intra- and post-operative data generated in the cochlear implantation process has steadily increased, partly due to technical innovations. To date, there is little evidence on the interplay of pre- and intraoperative factors that have an impact on hearing outcome. Furthermore, it is not possible to predict how hearing outcome will develop before or shortly after implantation. Predictive models represent a way to consider and combine a large amount of heterogeneous data (anamnestic, diagnostic and therapeutic). This is done using multi-modal machine learning techniques. This is a form of artificial intelligence (AI) in medicine. Aim: The development and validation of a predictive model for patients with cochlear implants is the aim of this project, which is conducted by the ENT Clinic of the University Medical Center Essen in collaboration with the Institute for Artificial Intelligence in Medicine (IKIM) and the Department of Mathematics and Computer Science, University of Marburg. By developing and validating a predictive model for patients with cochlear implants, we aim to identify factors that influence the hearing outcome of cochlear implantation. Thus, an objective indication and preoperative prediction of the postoperative hearing outcome should be possible. In this way, we hope to improve the indication with optimization of hearing in patients who receive a cochlear implant. Material and Methods: To achieve the project goal, a work program was designed, which is divided into 5 work packages. In work package 1 the digitization of the cochlear implant process pre-, intra- and postoperatively is planned. This serves as a prerequisite for work package 2, in which an analysis of individual influencing factors is calculated. In work package 3 an explainable prediction model will be developed, which learns to perform a preoperative selection of patients on the basis of retrospective pre- intra- and postoperative data as well as hearing success. The prediction model will be applied and evaluated in our center as well as in cooperating centers in work package 4 over a period of 24 months in a multicenter setting. In work package 5, the prediction model will be made available to CI centers in Germany. Outlook: With this innovative project at the interface between medicine, IT and informatics, independence from subjective indication is achieved by drawing on objective criteria and the diagnostic process is accelerated. The great innovation potential of the present application represents an opportunity to improve the cochlear implantation process.

• DFG Programme: Research Grants
• Co-Investigator: Professorin Dr. Diana Arweiler-Harbeck