A decreasing perinatal mortality in developed countries has reciprocally increased the prevalence of Cerebral Palsy to 2.11 per 1,000 live births. Within this population, poor muscle control, impaired balance and spasticity cause significant motor impairments. To improve locomotion, a surgery is the key part of treatment, whose success strongly depends on a correct surgery indication. However, the selection of the convenient surgery is very difficult and subjective also for expert surgeons. On the other hand, current motion capture systems can accurately measure kinematic, kinetic and spatio-temporal parameters of gait (so-called 3D motion data). A manual analysis of such complex data is practically impossible in huge data volumes. Within this project, we propose to develop a prototype software application for automatic analysis and evaluation of 3D motion data using sophisticated technologies employing artificial intelligence. In particular, we plan to utilize current advances in deep neural networks to learn a locomotion model from 3D motion data of thousands of gait cycles of more than 500 patients operated in the past. The learned model will be used to find formerly-operated patients with the most similar walking patterns. By considering the treatment outcome of these known patients, we could recommend the most suitable surgery option for a new examined patient. We assume that a sophisticated similarity analysis of a huge amount of 3D motion data can be the critical factor in distinguishing patients with a good response from patients with a poor response to a specific surgery, which is not possible to do in any other way at the present time. This project contributes to the healthcare of population (especially children aged 4–18) suffering from Cerebral Palsy and inherently requires a cooperation of experienced medical (MED MU) and technical (FI MU) institutions.

Publications