Empowering Foot Health: Harnessing the Adaptive Weighted Sub-Gradient Convolutional Neural Network for Diabetic Foot...

Area of Science:

• Medical Informatics
• Artificial Intelligence
• Computer Science

Background:

• Diabetic foot ulcer (DFU) presents a significant global health challenge, necessitating prompt and accurate diagnosis to prevent severe complications like amputation.
• Clinical assessment of DFU is often complex and time-consuming.
• Deep learning (DL) offers promising capabilities for improving DFU classification accuracy.

Purpose of the Study:

• To introduce and evaluate an Adaptive Weighted Sub-gradient Convolutional Neural Network (AWSg-CNN) for enhanced DFU classification.
• To improve the accuracy and efficiency of DFU diagnosis through advanced deep learning techniques.

Main Methods:

• The study employed a DFUC dataset, initiating with data pre-processing to remove inconsistencies and enhance data quality.
• A novel AWSg-CNN model was utilized for DFU classification, incorporating random initialization of weights (RIW) and log softmax with Adaptive Sub-gradient Optimizer (ASGO).
• RIW facilitates feature space learning, log softmax prevents gradient underflow, and ASGO controls gradient steps for optimized learning rates.

Main Results:

• The proposed AWSg-CNN method demonstrated superior performance in classifying DFU compared to traditional methods.
• The system achieved high accuracy, recall, F1-score, and precision, confirming its effectiveness.
• Results were visualized via a web interface using HTML, CSS, and Flask frameworks.

Conclusions:

• The AWSg-CNN model offers a robust and accurate solution for DFU classification.
• This deep learning approach has the potential to significantly aid in the early detection and management of diabetic foot ulcers.
• The developed system provides a practical tool for healthcare professionals in managing DFU patients.