Identification of kinematic motifs in crouch gait of children with cerebral palsy

Abstract Crouch gait is among the most disabling gait deviations in cerebral palsy (CP). Powered exoskeletons can supply the extensor thrust orthoses cannot, but their control depends on repeatable kinematic patterns, and crouch gait is perceived as irregular. Five children with CP (8–14 years; Gross Motor Function Classification System [GMFCS] I–III) walked at self-selected speed wearing bilateral shank inertial measurement units, knee goniometers and insole pressure sensors at the heel and forefoot. Signals were filtered with a two-state Kalman filter, adaptively segmented into gait cycles using plantar-pressure events (heel or toe strike), and time-normalised to 101 points. A kinematic motif was defined as a limb’s ensemble waveform and validated using the coefficient of variation, ICC(3,1), dynamic time warping, autocorrelation and 1,000-iteration permutation testing; symmetry was quantified with the Symmetry Index. Repeatability tracked the signal measured rather than functional severity: mean ICC(3,1) was 0.76 for forefoot loading, 0.61 for heel loading and 0.29 for knee kinematics. Full criteria were met in 12 of 30 limb–signal combinations, 12 met one or two, and 6 met none. Five of the six failures were knee channels, also the only channels indistinguishable from a permuted null, yet the knee met all criteria bilaterally in the least affected participant. These preliminary findings establish feasibility in a heterogeneous cohort ( n = 5); no gait-phase classifier was trained or evaluated here.