Full-field burn depth detection based on near-infrared hyperspectral imaging and ensemble regression

Abstract

The accurate and instant diagnosis of burn severity is always the key point of optimal wound management and clinical treatment. However, the accuracy of burn depth assessment is low via visual inspection and lacks a quantitative measurement. In this work, a full-field burn depth detection system is proposed using the near-infrared hyperspectral imaging with the ensemble regression. The rotational feature subspace ensemble regression is introduced to establish a complex regression model between the hyperspectral imaging data and the burn depth. By the in vivo measurement of a porcine model, the method can get the average relative error about 7% for the burn depth measurement, which demonstrates that the proposed method can perform an accurate full-field assessment of burn depth and provide more practical references for clinicians.