Diabetic Foot-Ulcer Detection from Digital Images using ResNet Variants and Particle Swarm Optimization

The occurrence rate of diabetes is increasing particularly in low- and middle-income countries due to various causes and it may cause mild to harsh illness. In order to control his harshness, early diagnosis and treatment is crucial. Diabetic Foot-Ulcer (DFU) is a major issue of diabetes and appropriate detection is necessary for the treatment. This work aims to detect the DFU from the digital-images using deep-learning (DL) technique. The stages in DL-scheme based DFU detection includes; digital-image collection and resizing, feature mining using a chosen ResNet-model, features selection using Particle Swarm Optimization (PSO), fused features vector generation using PSO selected features, and classification and verification. In this work, the digital-image based DFU detection is executed using 1050 images (525 healthy and 525 DFU) and the proposed investigation is executed using Python-software. The classification result of this study confirms that the fused-features based DFU detection offered >99% accuracy with the SoftMax-classifier. This confirms that the implemented scheme can be chosen to examine the clinically collected digital-images having the DFU.