Related Experiment Videos
Explainable AI-based clinical decision support system for flatfoot classification using deep learning
V Harsha Vardhan1, S G Rahul1, T M Amirthalakshmi1
1Department of Electronics and Communication Engineering, Amrita School of Engineering, Amrita Vishwa Vidyapeetham, Chennai, India.
Frontiers in Digital Health
|August 13, 2026
Summary
This study compared machine learning and deep learning for flatfoot (pes planus) detection. Deep learning models, particularly ResNet101V2, showed superior performance, achieving over 92% accuracy in classifying flatfoot deformities.
Area of Science:
- Biomedical Engineering
- Computer Science
- Orthopedics
Background:
- Flatfoot (pes planus) is a foot deformity characterized by reduced medial longitudinal arch height.
- This condition can lead to significant postural and locomotion issues.
Purpose of the Study:
- To compare the efficacy of handcrafted feature-based machine learning models and deep learning models for flatfoot classification.
- To evaluate the explainability of the models using SHAP and Grad-CAM techniques.
Main Methods:
- Comparative analysis of machine learning (LBP with Random Forest, Decision Tree, Logistic Regression) and deep learning (InceptionResNetV2, ResNet101V2, DenseNet201, DenseNet169, InceptionV3, Xception) models.
- Monte Carlo cross-validation with subject-wise splitting was utilized.
- SHAP and Grad-CAM were employed for model interpretability.
Main Results:
- Random Forest with LBP features achieved 70% accuracy.
- ResNet101V2-RMSprop demonstrated a mean accuracy of 92.86% ± 2.82%.
- Deep features combined with a Decision Tree model yielded 97% accuracy, 95% recall, and 100% specificity.
Conclusions:
- Deep learning models, especially ResNet101V2 and a hybrid deep feature/Decision Tree approach, significantly outperform traditional machine learning for flatfoot classification.
- Explainability methods confirmed that models focus on clinically relevant anatomical regions, supporting their utility in clinical screening for pes planus.