Three major one layer tissue models (Modified Beer-Lambert,1 Jacques 1999,2 Pilon 20093) are compared to Monte Carlo simulated diffuse reflectance spectra and measured tissue phantom spectra with known ground truth. These ground truth values were obtained using inverse adding doubling and absorbance measurements and validated using a phantom with known ground truth (BioPixs). Finally, a two layer model (Pilon 2009) was evaluated against Monte Carlo simulations and used to analyse skin reflectance data (NIST4). These models were compared on goodness of fit and parameter extraction accuracy. It was found that the Pilon 2009 one layer model performed best against Monte Carlo simulations and phantom measurements, however the Pilon 2009 two layer model had significant regions of inaccuracy. These inaccurate regions correspond to circumstances where the epidermal layer has significant thickness and melanin content, while the dermal layer has low fraction of blood meaning that the haemoglobin impact is “masked”. The extraction of parameters from the NIST skin dataset using this model returns values that do not correspond well to literature values suggesting that many of these spectra lie within an inaccurate region or indicates oversimplification of the tissue modelling. This suggests both Pilon 2009 and Jacques 1999 are suitable for modelling tissue that can be approximated as a single, homogeneous, semi-infinite slab, however the Pilon 2009 two layer model is not yet effective when encountering empirical data.
|