An Enhanced Diabetic Foot Ulcer Classification Approach Using GLCM and Deep Convolution Neural Network

Hussein A. Ismael, College of Information Technology, University of Babylon, Hillah, Babil,Follow
Nabeel H. Al-A’araji, College of Information Technology, University of Babylon, Hillah
Baheja khudair shukur, College of Computer Science & Information Technology, University of Kerbala, Kerbala, Iraq

Abstract

Diabetic Foot Ulcers (DFU) are considered to be a common complication of diabetes, usually resulting in the amputation of lower extremities. Therefore, diagnosing this disease at an early stage is necessary to avoid the accompanying treatment approach, and this results in a significant cost reduction for the patient. To achieve an early diagnosis of this disease, we need to classify a patient's skin as normal or abnormal. A classification process relies heavily on the extracted features. So, we proposed a new technique called CNN_GLCMNet for feature extraction. This technique relies on Convolution Neural Network (CNN) and the Gray-Level Co-Occurrence Matrix (GLCM) techniques to mine abstract features and second-order statistical texture features. Also, Singular Value Decomposition (SVD) is applied to reduce the dimensionality of the obtained features that result from CNN, Next, the GLCM method is applied to extract second-order statistical texture features. Then, these two kinds of features (abstract features and statistical features) are combined and used as input for the classifier. Two classification mechanisms have been adopted in the classification of images into normal and abnormal skin. First, the Deep Neural Network (DNN) classifier achieves the following performance evaluation metrics (accuracy 97.43%, recall 97.25%, specificity 97.59%, precision 97.53%, f1-score 97.38%). Second, the Support Vector Machine (SVM) classifier achieves the following performance evaluation metrics (accuracy 96.93%, recall 96.99%, specificity 96.94%, precision 96.76%, f1-score 96.85%). Since both classifiers have been validated against the DFU dataset using 10-fold cross-validation. The DNN classifiers with our new feature extraction technique achieve better results in terms of accuracy, specificity, precision, recall, and f1-score than in previous work. Furthermore, a comparison of DNN and SVM classifiers finds that DNN gives a better result according to performance metrics.

Recommended Citation

Ismael, Hussein A.; Al-A’araji, Nabeel H.; and shukur, Baheja khudair (2022) "An Enhanced Diabetic Foot Ulcer Classification Approach Using GLCM and Deep Convolution Neural Network," Karbala International Journal of Modern Science: Vol. 8 : Iss. 4 , Article 11.
Available at: https://doi.org/10.33640/2405-609X.3268