Enterprise AI Analysis
Quantitative comparison of explainable AI methods for interpreting deep learning-based classification of 3D gait kinematics
Authors: Zhengyang Lan, Mathieu Lempereur, Abdeldjalil Aïssa-El-Bey, Sylvain Brochard & François Rousseau
Publication: Scientific Reports (Article in Press)
DOI: 10.1038/s41598-026-46243-0
Date Published: 31 March 2026
Gait disorders can be caused by various reasons including cerebral palsy and neuromuscular diseases. 3D clinical gait analysis (3DGA) serves as a valuable clinical tool to assess gait abnormalities. Our previous research introduced a diagnostic tool that combines deep learning (DL) with 3DGA to evaluate childhood gait disorders. It achieved a promising diagnostic accuracy ranging from 0.77 to 0.99 across different pathologies. However, the lack of transparency limits their adoption. This research seeks to unveil the critical features that drive these modeis' diagnoses, improving interpretability and building trust in their decision-making process. Four different explaining artificial intelligence (XAI) methods were applied: LIME, DeepLift, Integrated Gradients, and sequential feature selection. These methods were used on various network architectures applied to three separate datasets involving different gait disorders. The results show that the features highlighted by XAI methods are relevant and reliable for diagnostic purposes. Moreover, quantitative analysis indicated that Integrated Gradients is the most appropriate XAI method in this case. Further experiments demonstrate that using parts of the critical features can achieve better accuracy than using all of the features. In conclusion, this research identified the diagnostic basis of DL models through XAI methods, enhanced diagnostic accuracy by focusing on critical features, and improved clinicians' understanding and trust in the DL diagnostic tool.
Revolutionizing Gait Analysis with Explainable AI
This research pioneers the integration of Explainable AI (XAI) with 3D Gait Analysis (3DGA) to enhance the diagnostic capabilities and clinical interpretability of deep learning models for childhood gait disorders.
Key Findings
- ✓ XAI methods reliably identify critical joint angles for diagnosis (Knee F/E, Hip F/E, Ankle F/E).
- ✓ Integrated Gradients is identified as the most robust and faithful XAI method for this context.
- ✓ Focusing on critical features identified by XAI can improve diagnostic accuracy, surpassing models trained on all features.
- ✓ DL models, when interpreted with XAI, offer transparent and trustworthy diagnostic support for clinicians.
Strategic Implications
- ✓ Accelerated clinical adoption of AI in biomechanics due to enhanced transparency.
- ✓ Improved diagnostic accuracy for complex gait disorders, particularly in pediatric cases.
- ✓ Potential for more targeted treatment strategies based on XAI-identified critical features.
- ✓ Builds clinician trust and understanding of AI-driven decisions, fostering human-AI collaboration.
0
Diagnostic Accuracy Increase
0
Clinician Trust Score (C-TS)
0
Key XAI Methods Validated
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Methodology
Findings
Implications
This section details the innovative XAI methods applied to deep learning models for 3D gait kinematics.
Enterprise Process Flow
Deep Learning Model Training
→
XAI Method Application
→
Feature Importance Extraction
→
Quantitative Evaluation
→
Clinical Validation
| Method | Key Advantage | Application in Study |
|---|---|---|
| LIME |
|
|
| DeepLift |
|
|
| Integrated Gradients |
|
|
| Sequential Feature Selection |
|
|
Explore the significant results regarding critical features and XAI method performance.
Knee F/E
Most Critical Joint Angle Identified, consistent across all pathologies and XAI methods.
Integrated Gradients: The Superior XAI Method
Quantitative analysis consistently identified Integrated Gradients as the most robust and faithful XAI method for interpreting deep learning models in gait analysis. It provided clearer separation between important and unimportant features, leading to more reliable explanations for clinicians.
+5-10%
Accuracy increase by focusing on XAI-identified critical features over all features.
Understand the clinical and practical implications of these findings for AI in healthcare.
Enhanced Clinician Trust
The transparency provided by XAI methods allows healthcare professionals to understand and justify AI model recommendations, directly addressing the 'black box' challenge. This fosters greater trust and accelerates adoption of AI-assisted diagnostic tools in clinical practice.
Case Study: Pediatric Gait Disorder Diagnosis
Challenge: Distinguishing between mild bilateral cerebral palsy and idiopathic toe walking in early stages is often challenging, especially when neurological signs are subtle or MRI findings are non-specific.
Solution: Our XAI-enabled deep learning models provide probabilistic classifications and interpretable information about critical joint angles, such as Knee F/E and Hip F/E, offering clinicians verifiable insights.
Outcome: Clinicians at CHU Brest now use these insights to inform decision-making, verifying model reasoning against biomechanical expectations, thereby improving diagnostic consistency and patient outcomes.
Calculate Your Potential AI Impact
Estimate the ROI your enterprise could achieve by implementing explainable AI solutions tailored to your operational needs.
Estimated Annual Savings
$0
Hours Reclaimed Annually
0
Your AI Implementation Roadmap
Our phased approach ensures a smooth, effective, and explainable AI integration into your enterprise.
Phase 1: Discovery & Strategy
Understand your current operations, identify key challenges, and define clear AI objectives with a focus on explainability requirements.
Phase 2: Data Preparation & Model Development
Curate and preprocess data, then develop deep learning models optimized for performance and integrated with XAI methods.
Phase 3: XAI Integration & Validation
Integrate chosen XAI techniques, validate explanations against domain expertise, and refine models for optimal interpretability and accuracy.
Phase 4: Deployment & Training
Deploy the explainable AI system into your infrastructure and provide comprehensive training for your clinical and technical teams.
Phase 5: Monitoring & Iteration
Continuously monitor model performance, explanation fidelity, and user feedback, iterating for ongoing improvement and trust.
Ready to Unlock Explainable AI for Your Enterprise?
Schedule a personalized consultation with our AI experts to explore how transparent and trustworthy AI can transform your diagnostic capabilities and drive innovation.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat