Deep Learning-Based Classification and Feature Extraction for Predicting Pathogenesis of Foot Ulcers in Patients with...

Area of Science:

• Medical Imaging and Artificial Intelligence
• Diabetology and Podiatry

Background:

• Diabetes mellitus (DM) is a global health concern with significant mortality, often exacerbated by complications like diabetic foot ulcers (DFU).
• DFU can lead to severe outcomes, including amputation, highlighting the need for early detection and risk assessment.
• Lack of awareness regarding DFU risks contributes to patient mortality.

Purpose of the Study:

• To develop and evaluate a novel methodology for analyzing diabetic foot ulcer images to predict the risk of ulceration.
• To leverage deep learning techniques for early symptom detection and effective treatment planning for diabetic patients.

Main Methods:

• Collected and pre-processed a dataset of historical foot ulcer images and medical records from diabetic patients.
• Utilized a deep recurrent neural network (DRNN) for feature extraction from segmented image and numerical/text data.
• Employed a pre-trained fast convolutional neural network (PFCNN) with U++net for classifying abnormal diabetes foot images and predicting DFU risk.

Main Results:

• The proposed technique achieved a high accuracy of 99.32% in classifying normal versus abnormal diabetic foot ulcerations.
• Simulation results, including confusion matrices and ROC curves, demonstrated the effectiveness of the feature extraction and classification methods.
• Key performance parameters such as accuracy, precision, and area under the curve were analyzed.

Conclusions:

• The developed deep learning approach offers a promising, highly accurate method for assessing DFU risk through image analysis.
• Early prediction of DFU using this technique can significantly improve patient outcomes and potentially reduce amputation rates.
• The methodology provides a valuable tool for clinicians in managing diabetic foot complications.