Diabetic foot ulcer classification using an enhanced coordinate attention integrated ConvNext model.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Diabetes Management

Background:

• Diabetic foot ulcers (DFUs) are a serious diabetes complication, often leading to amputation.
• Current DFU management is costly and requires intensive monitoring, with limitations in pattern recognition for accurate classification.
• Existing methods struggle with identifying complex patterns and contextual correlations in DFU images.

Purpose of the Study:

• To develop an automated, deep learning-based approach for enhanced Diabetic Foot Ulcer (DFU) assessment using medical images.
• To improve the accuracy and efficiency of DFU investigation and recommendation processes.
• To overcome limitations in current DFU diagnostic methods regarding pattern recognition and contextual analysis.

Main Methods:

• Employed adaptive thresholding to enhance DFU image contrast and uniformity for improved feature extraction.
• Utilized a hybrid deep learning model combining ConvNeXt architecture with coordinate attention for DFU image classification.
• Incorporated coordinate attention to capture spatial information, enhancing the extraction of long-range dependency features.

Main Results:

• The developed model achieved a high classification accuracy of 97.16%.
• An F1-score of 0.97 was obtained, indicating robust performance in DFU identification.
• The attention-enhanced ConvNeXt model demonstrated effective representation of complex patterns in DFU images.

Conclusions:

• The proposed attention-enhanced deep learning model offers a promising automated solution for DFU assessment.
• This approach can expedite DFU investigation, leading to more optimal treatment recommendations.
• The findings suggest a significant advancement in leveraging AI for improved diabetes complication management.