Cerebral Palsy Children
Conditions
Keywords
CP Assesment, Explainable machine learning, Gait symmetry, Functional independence
Brief summary
The goal of this observational study is to develop and validate an AI-based prediction model for functional mobility and gait outcomes in children with cerebral palsy using low-cost clinical and gait data collected in rehabilitation settings in Pakistan.
Detailed description
Children with cerebral palsy (CP) commonly experience limitations in functional independence and mobility, which significantly affect participation and quality of life. Accurate assessment of these functional abilities is essential for rehabilitation planning, prognosis estimation, and monitoring treatment outcomes. However, conventional assessment methods largely depend on therapist observation and standardized clinical scales, which may be subjective, time-consuming, and less sensitive to complex interactions among clinical variables.
Interventions
OTHERAI-Based Functional Mobility and Gait Assessment
Participants will continue receiving their standard/routine physiotherapy rehabilitation program as prescribed by their treating therapist. The study will involve observational collection of clinical, functional, and gait-related data using standardized assessment tools, and AI-based video analysis. No additional therapeutic intervention will be administered specifically for research purposes.
Sponsors
Riphah International University
Study design
Observational model
COHORT
Time perspective
PROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
4 Years to 18 Years
Healthy volunteers
No
Inclusion criteria
* Age 4 to18 years * Diagnosed any motor type of cerebral palsy (spastic, dyskinetic, ataxic, mixed),) * GMFCS levels I -III (able to walk with or without an assistive device). * All participants must be able to ambulate at least 10 meters with or without an assistive device. * Capable of following simple verbal instructions. * Parental informed consent and child assent
Exclusion criteria
* Recent orthopedic or neurosurgical interventions (\<6 months). * Uncontrolled seizures affecting gait. * Non-ambulatory (GMFCS IV-V) or cognitive impairments preventing cooperation.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| GMFM-88 | Baseline to 6 months followup | GMFM (Gross Motor Function Measure) Reliability: Excellent. Internal consistency Cronbach's α \~0.997-1.00; intra- and inter-rater ICC \~0.994-0.999 (both GMFM-88 \& GMFM-66) Validity: Construct and concurrent validity supported by strong correlations with related motor function classifications (e.g., GMFCS, PEDI mobility) |
| Markerless Gait Analysis | Baseline to 6 months | Gait videos will be processed using a validated markerless pose estimation framework . Spatiotemporal and kinematic gait parameters will be extracted, including but not limited to: * Step length symmetry * Cadence * Stride time variability * Joint angle trajectories * Temporal asymmetry indices |
| Edinburgh visual gait scale (EVGS) | Baseline to 6 Months | Edinburgh visual gait scale (EVGS) EVGS can be a supportive tool that adds quantitative data instead of only qualitative assessment to a video only gait evaluation. Interobserver agreement is 60-90% and Kappa values are 0.18-0.85 for the 17 items in EVGS. Reliability is higher for distal segments (foot/ankle/knee 63-90%; trunk/pelvis/hip 60-76%). Agreement between EVGS and 3DGA is 52-73%. |
| WeeFIM (Functional Independence Measure for Children) | Baseline to 6 months | WeeFIM (Functional Independence Measure for Children) Reliability: High internal consistency and ICCs (motor and cognitive scales) \~0.91-0.98 in children with cerebral palsy Validity: Construct and external validity supported (scale fits Rasch model expectations and correlates with related developmental measures) |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Feasibility of markerless video-based gait analysis for routine physiotherapy assessment in low-resource clinical settings | Through study completion, 12 month | Feasibility will be evaluated using: (1) video acquisition success rate; (2) data processing completion rate; (3) time required for analysis; and (4) clinician usability and interpretability feedback, obtained through structured questionnaires and semi-structured interviews with practicing physiotherapists regarding SHAP-based model outputs. |
| Predictive accuracy of the machine learning model for gait and motor function | At model validation ,post data collection ,6 month | Model performance will be evaluated using RMSE, MAE, and R² for continuous outcomes. Explainability will be assessed using SHAP values, with examination of feature importance consistency across cross-validation folds. based on gait and motor function data collected from participants with ambulatory cerebral palsy. |
| Robustness of model predictive performance and change in functional and gait outcomes across heterogeneous therapy exposure contexts | Baseline to 6-month follow-up | The Phase 2 trained model will be applied to 6-month follow-up data without retraining. Predictive performance (RMSE, MAE, R²) will be compared across therapy exposure subgroups (regular vs. irregular/no physiotherapy). Temporal prognosis will additionally be characterized using change from baseline in GMFM-66, WeeFIM, and EVGS, alongside gait symmetry indices derived from video analysis. |
Countries
Pakistan
Contacts
CONTACTQamar Mehmood, Phd Rehab
PRINCIPAL_INVESTIGATORSidra Ghias, PhD* Rehab
Riphah International university Isalambad
Outcome results
None listed