Efficient Detection and Classification of Diabetic Foot Ulcer Tissue using PSO Technique

Abstract

Assessment of a diabetic wound is very much important to determine the healing status. Foot ulcer is most commonly observed problem of diabetic patients. A diabetic wound is observed for approximately 15 per cent of diabetic patients. Diabetic wound is a major concern of diabetes mellitus. The foot ulcer is the very much harm full problem related to diabetes mellitus. Here particle swarm optimization (PSO) based optimization technique is used for segmentation of diabetic wounds and classifying into three types of tissues i.e. granulation, necrotic and slough. After the segmentation the different textural features are extracted through Gray Level Co-occurrence Matrix (GLCM). All these features were then fed to two different classifiers, Naive bayes and Hoeffding tree for classifying the tissue types. The experimental results showed that the classification accuracy, sensitivity, specificity are 90.90%, 100%, 87.5% by Naive bayes, and 81.81%, 100%, 77.7% by Hoeffding tree respectively. Hence the PSO optimization techniques along with Naive bayes classifier could be used for the effective segmentation of diabetic wound images.