A hybrid markerless approach for 2D gait analysis: application to gait of children with cerebral palsy

A 2D hybrid markerless technique is proposed to estimate the lower limb joint kinematics in children with cerebral palsy (CP). The markerless methodology presented is defined as hybrid since it uses garments as “segmental markers”, specifically high-cut underwear which contributes in defining the pelvis segment and ankle socks for the definition of the foot segment. Six CP children participated in this study and five trials were recorded for each subject. The proposed technique requires the following steps: anthropometric measurements, video acquisition (Figure a), segmentation (Figure b), skeletonization (Figure c), thresholding and edge detection (Figure d), determining intersection points (Figure e) and fitting reference axes (Figure f). Validation was done by comparing the lower limb joint kinematics estimated by the hybrid approach with those obtained using a marker-based system. The average angular offsets were in the range of 2.9- 4.5 for ankle plantar/dorsi-flexion (α), 0.3-2.1 for knee flexion/extension (β) and 0.4-2.4 for hip flexion/extension (γ). The average RMSD results were in the range of 2.8-6.0 for the (α), 3.9-8.5 (β) and 3.9-7.6 and (γ) angles. RMSD values were in general within the intra-subject variability. The proposed technique can be a possible economical alternative for estimating the lower limb joint kinematics of children with CP.