DFU-GCNet: a global context-enhanced inception network for robust and interpretable diabetic foot ulcer...

Area of Science:

• Medical imaging
• Artificial intelligence in healthcare
• Diabetic wound management

Background:

• Diabetic foot ulcers (DFUs) are a leading cause of lower extremity amputations.
• Accurate and timely diagnosis of DFUs is critical for effective treatment.
• Existing deep learning models for DFU detection face challenges with multi-scale lesions and lack clinical transparency.

Purpose of the Study:

• To develop a robust and interpretable deep learning model for DFU classification.
• To address limitations in current DFU detection methods, including scale variability and opaque decision-making.
• To enhance clinician trust in automated diagnostic tools for diabetic foot conditions.

Main Methods:

• Introduction of DFU-GCNet, an architecture combining inception modules and global context blocks.
• Extraction of multi-scale features and modeling of spatial dependencies for improved pathology detection.
• Integration of explainable AI techniques (GradCAM++, LIME, SHAP) for clinical transparency and validation using the Kaggle DFU dataset.

Main Results:

• DFU-GCNet achieved a classification accuracy of 97.16%, with an F1-score of 0.9715 and a Matthews correlation coefficient of 0.9437.
• The model demonstrated superior performance compared to established baselines like VGG16 and EfficientNet.
• Explainable AI methods confirmed that the network focuses on clinically relevant wound boundaries.

Conclusions:

• DFU-GCNet serves as a highly reliable automated screening instrument for diabetic foot ulcers.
• The model's interpretability fosters greater clinical trust and facilitates adoption in healthcare settings.
• This approach advances the potential of AI in managing diabetic complications and preventing amputations.