Review of machine learning for optical imaging of burn wound severity assessment

• Published in: Journal of biomedical optics 2024-02, Vol.29 (2), p.020901-020901
• Main Authors: Wilson, Robert H., Rowland, Rebecca, Kennedy, Gordon T., Campbell, Chris, Joe, Victor C., Chin, Theresa L., Burmeister, David M., Christy, Robert J., Durkin, Anthony J.
• Format: Article
• Language: English
• Subjects: Algorithms; Burns - diagnostic imaging; Humans; Machine Learning; Optical Imaging - methods; Review Papers; Skin
• ISSN: 1083-3668; 1560-2281; 1560-2281
• DOI: 10.1117/1.JBO.29.2.020901

Over the past decade, machine learning (ML) algorithms have rapidly become much more widespread for numerous biomedical applications, including the diagnosis and categorization of disease and injury. Here, we seek to characterize the recent growth of ML techniques that use imaging data to classify burn wound severity and report on the accuracies of different approaches. To this end, we present a comprehensive literature review of preclinical and clinical studies using ML techniques to classify the severity of burn wounds. The majority of these reports used digital color photographs as input data to the classification algorithms, but recently there has been an increasing prevalence of the use of ML approaches using input data from more advanced optical imaging modalities (e.g., multispectral and hyperspectral imaging, optical coherence tomography), in addition to multimodal techniques. The classification accuracy of the different methods is reported; it typically ranges from to 90% relative to the current gold standard of clinical judgment. The field would benefit from systematic analysis of the effects of different input data modalities, training/testing sets, and ML classifiers on the reported accuracy. Despite this current limitation, ML-based algorithms show significant promise for assisting in objectively classifying burn wound severity.