Automated Infection Detection in Diabetic Foot Ulcer Images Using CNN and ELM Models

Severe consequences of diabetes encompass diabetic foot ulcers (DFUs), which may result in limb or foot amputation and are a significant contributor to mortality, disability, and healthcare costs globally. Enhancing healing outcomes and reducing the likelihood of gangrene and amputation in diabetic foot ulcers necessitates prompt infection identification with the recognition of ischemia, attributable to diminished blood flow due to chronic diabetes. This study proposes a novel method for detecting infections in diabetic foot ulcers by a hybrid classification system. Subsequent to the fine-tuning of the dataset using precise edge detection, it was equilibrated by incorporating additional data through data augmentation techniques. The GLCM approaches were employed for feature extraction as they encapsulate essential second-order statistical texture information vital for effective categorization. A CNN-ELM model was developed, integrating the automated feature extraction skills of CNN with the efficiency and precision of ELM. The model discerned six stages of DFU development, achieving an exceptional classification accuracy of 98.31% and an F1-score of 98.87%, despite the constraints of limited training data. To enhance early identification, treatment outcomes, and the need for amputation, these findings indicate that the proposed method provides a rapid, precise, and efficient tool for detecting infections in diabetic foot ulcers.