Abstract
Hyperspectral imaging (HSI) is an emerging technique that is suitable for tissue oxygen saturation (StO2) assessment. In the past, different ranges of wavelengths and different Beer Lambert law models were employed for the assessment. However, the reasons why these spectral ranges and models were chosen remain unknown. The aim of the present paper is to elucidate why subsets of spectral data and modified Beer Lambert models are more suitable than others. We used four different Beer Lambert models under various subset spectral regions within 450–850 nm to deduce StO2 of a human palm under two illumination conditions. Experimental results show that the subset spectral region between 516 and 580 nm is more suitable than other subset regions to assess StO2 and that the modified Beer Lambert model using three chromophores can give the smallest fitting error. This work suggests that this subset of spectra and the modified Beer Lambert model are more appropriate for StO2 assessment using HSI.