Comparison of photometric stereo and spectral analysis for visualization and assessment of burn injury from hyperspectral imaging

摘要

Burn wounds resulting from thermal insult to the skin are typically classified according to varying depth and therefore require differential levels of medical intervention. In this paper, two methods are proposed for assessing burn injury. The two methods compared are photometric stereo (PS) and spectral analysis. Firstly, PS represents a robust topography recovery algorithm that is implemented to reconstruct the burn and normal skin tissue from multiple hyperspectral images under different illumination conditions. This enabled the visualization of a 3D skin depth map which is used to assess the burn degree. Next, the hyperspectral measurement data of the skin are analyzed to assess partial thickness thermal injury with functional correlation through hemodynamic parameters related to tissue perfusion and oxygen delivery. Two dimensional principle component analysis (2DPCA) is used for noise reduction towards extracting features from the hyperspectral images within the wavelength range from 375 nm to 1050 nm. This is followed by applying the spectral analysis algorithm to calculate oxygen saturation fraction and concentration of total hemoglobin, where each parameter provided a biomarker of injured tissue. The two methods yielded alternative indicators for burn assessment that could be correlated with each other. Specifically, the spectral measurement result could be used as a reference value for the physical skin site depth map.