A New Pes Planus Automatic Diagnosis Method: ViT-OELM Hybrid Modeling

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Deep Learning

Background:

• Pes planus, or flat feet, is a common foot condition characterized by flattened soles.
• Traditional diagnostic methods can be subjective and time-consuming.
• Deep learning offers potential for automated and objective diagnosis of medical conditions.

Purpose of the Study:

• To propose and evaluate a novel Vision Transformer-Optimized Extreme Learning Machine (ViT-OELM) hybrid model for the automated diagnosis of pes planus.
• To compare the performance of the proposed ViT-OELM model against existing methods using the same dataset.
• To leverage deep learning for enhanced accuracy in identifying pes planus from foot images.

Main Methods:

• Utilized an openly available Kaggle dataset of pes planus images.
• Developed a ViT-OELM hybrid deep learning architecture incorporating an attention mechanism for feature extraction.
• Employed an Optimum Extreme Learning Machine (OELM) classifier with extracted features for binary classification (pes planus vs. not pes planus).

Main Results:

• The ViT-OELM model achieved high performance metrics in binary classification.
• Achieved an accuracy of 98.04%, recall of 98.04%, precision of 98.05%, and an F-1 score of 98.03%.
• Demonstrated superior performance compared to other studies using the same pes planus dataset.

Conclusions:

• The proposed ViT-OELM hybrid model is effective for automated pes planus diagnosis.
• The model's high accuracy and performance metrics indicate its clinical potential.
• This deep learning approach offers a promising advancement in the objective assessment of flat feet.