Clinical Assessment of Diabetic Foot Ulcers Using GWO-CNN based Hyperspectral Image Processing Approach
T. Arumuga Maria Devi, R. Hepzibai
Abstract
Diabetes Mellitus has turned out to be a complicated disease and as of 2016 one out of eleven humans suffer from this disease leading to Diabetic Foot Ulcers (DFU). When not treated DFUs lead to amputation and in this work, a novel image processing method is proposed for the efficient assessment and classification of DFU images. Initially, pre-processing is done by cascaded fuzzy filter followed by nonlinear partial differential equation (NPDE) based segmentation that segments the foot ulcer regions. Consequently, the local binary pattern (LBP) is employed to extract the useful features. Then the proposed hybrid Grey Wolf Optimization-Convolutional Neural Network (GWO-CNN) model uses these features to identify the DFU regions. The performance evaluation is done by the estimation of the performance metrics and the results are compared with existing algorithms indicating the efficacy of the proposed technique. The obtained results reveal that the proposed work generates an accuracy of 98.5% with a reduced error percentage of 1.4%.