LIVE-GaitNeuroKids: A multimodal and Longitudinal Gait dataset with Inertial, Video and surface Electromyography in Neuromotor conditions and typically developing children

Precise characterization of gait patterns is essential for the clinical assessment of motor disorders, such as cerebral palsy (CP) and idiopathic toe walking (ITW). While laboratory-based motion capture remains the gold standard, its operational complexity often limits its use in daily clinical practice. In this work, we present a comprehensive, multi-modal dataset of pediatric gait comprising 59 participants: 22 with CP, 12 with ITW, and 25 typically developing (TD) controls. The dataset provides synchronized time-series of 3D kinematics, derived from a body-worn inertial measurement unit (IMU) system (Xsens™), and four-channel surface electromyography (sEMG) (Delsys Trigno™) targeting the tibialis anterior and gastrocnemius muscles. Additionally, regular video recordings from three views (front left, front right, and sagittal) are provided as visual ground truth. A distinguishing feature of this repository is the inclusion of pre- and post-treatment sessions for pathological cohorts, enabling the evaluation of clinical interventions over time. The records include (i) raw acquisition files, (ii) processed clinical joint angles and sEMG signals, and (iii) segmented gait cycles time-aligned across wearable sensors. This dataset facilitates the development of automated diagnostic algorithms and digital gait biomarkers, bridging the gap between laboratory-based gait analysis and portable monitoring in pediatric clinical settings.