Wound image evaluation with machine learning

• Published in: Neurocomputing (Amsterdam) Vol. 164; pp. 112 - 122
• Main Authors: Veredas, Francisco J., Luque-Baena, Rafael M., Martín-Santos, Francisco J., Morilla-Herrera, Juan C., Morente, Laura
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 21.09.2015
• Subjects: Computer-aided diagnosis; Machine vision; Medical imaging; Wound evaluation; Medical imaging; Machine vision; Computer-aided diagnosis; Wound evaluation
• ISSN: 0925-2312; 1872-8286
• DOI: 10.1016/j.neucom.2014.12.091

A pressure ulcer is a clinical pathology of localized damage to the skin and underlying tissue caused by pressure, shear or friction. Diagnosis, care and treatment of pressure ulcers can result in extremely expensive costs for health systems. A reliable diagnosis supported by precise wound evaluation is crucial in order to succeed on the treatment decision and, in some cases, to save the patient׳s life. However, current clinical evaluation procedures, focused mainly on visual inspection, do not seem to be accurate enough to accomplish this important task. This paper presents a computer-vision approach based on image processing algorithms and supervised learning techniques to help detect and classify wound tissue types that play an important role in wound diagnosis. The system proposed involves the use of the k-means clustering algorithm for image segmentation and compares three different machine learning approaches—neural networks, support vector machines and random forest decision trees—to classify effectively each segmented region as the appropriate tissue type. Feature selection based on a wrapper approach with recursive feature elimination is shown to be effective in keeping the efficacy of the classifiers up and significantly reducing the number of necessary predictors. Results obtained show high performance rates from classifiers based on fitted neural networks, random forest models and support vector machines (overall accuracy on a testing set [95% CI], respectively: 81.87% [80.03%, 83.61%]; 87.37% [85.76%, 88.86%]; 88.08% [86.51%, 89.53%]), with significant differences found between the three machine learning approaches. This study seeks to provide, using standard classification algorithms, a consistent and robust methodological framework as a basis for the development of reliable computational systems to support ulcer diagnosis.