Towards a Model-based Digital Twin for Gait Analysis in Parkinson-Related Assessment

Abstract Step-width variation and foot-placement are relevant measures for Parkinson-related gait assessment, but are difficult to derive robustly from compact wearable sensor setups. Previous work within the same project used bodyworn magnetoelectric sensors to estimate step width from relative distance measurements, but did not provide a continuous kinematic body representation. This work presents a work-in-progress transition toward a model-based, digitaltwin- oriented framework for inertial measurement unit (IMU)- based gait analysis. The framework combines a modular processing chain, a hierarchical kinematic skeleton model, and a planned state-space estimation approach for recursive pose reconstruction from accelerometer and gyroscope signals. The current implementation comprises the individualized kinematic model, recording-specific marker alignment, and its integration into a shared simulation and processing framework. The kinematic model is individualized through patient-specific segment lengths and recording-specific marker offsets, while dynamic updating through wearable IMU signals is planned as a subsequent step. The ensuing development steps also include recursive state-space estimation and validated derivation of gait parameters.