FIGURE 2 . (a) 3D LC-OCT image of human skin in vivo , represented in slice view and (b) averaged intensity profileR (z ), showing mean linear regression fit of each skin layer (epidermis – red, dermis – blue) and a corresponding pairs of observables (\(\rho^{\text{epi}},\ µ_{\text{eff}}^{\text{epi}}\) and\(\rho^{\text{der}},\ µ_{\text{eff}}^{\text{der}}\)). Dermal layer parameter \(\rho^{\text{der}}\) deduced from the intercept with basal membrane (z = ~150 µm) and corrected from epidermal layer attenuation.
Then, a linear regression fit was applied separately to two parts of the intensity profile, corresponding to the epidermal and dermal layers (Figure 2(b)). Fitting areas were delineated manually for each layer, considering mostly linear parts of attenuation slope before the background of multiply scattered light becomes dominant and changes the slope.[13] µ eff parameter of each layer (\(µ_{\text{eff}}^{\text{epi}}\) for epidermis and\(µ_{\text{eff}}^{\text{der}}\) for dermis) was calculated as half of each linear fit slope. Epidermal ρepi parameter corresponding to the intercept of epidermal linear fit with depthz = 0, which is the interface between the LC-OCT probe glass surface and the skin surface. For the dermis,ρder was obtained from\(\rho^{\text{der}}e^{-2µ_{\text{eff}}^{\text{epi}}\text{δz}}\) (that is an intercept with between layer interface – basal membrane) by dividing it by correction factor\(\ e^{-2µ_{\text{eff}}^{\text{epi}}\text{δz}}\) (or by multiplying it with correction factor\(e^{2µ_{\text{eff}}^{\text{epi}}\text{δz}}\)) to compensate the attenuation by the epidermal layer of thickness δz (Figure 2(b)).
Resulting experimental observables of each layer were then mapped to the model, described by Equations 2 and 3, using the following expression proposed by Jacques et al ., considering µ s>>µ a:[13]
\(\frac{µ_{\text{eff}}}{\rho}=\ \frac{\text{aG}}{b\Delta z}\) (4)
Such expression has the advantage of being independent fromµ s and depends only on wavelength, NA andg . With our experimental values λ = 800 nm and NA = 0.5 , the anisotropy factor g of the corresponding layer was then retrieved as it is a function of µ eff/ρ (as demonstrated in the works of Jacques and Waszczuk).[13,16] After retrievingg (λ 800) parameter,µ s(λ 800) parameter can be calculated using Equations 2 and 3.
For each 3D image, a linear fit of each layer was repeated more than 100 times by scanning the fit limits by 10 µm with 1 µm step to estimate the variability of the fitting depending on the range chosen. Then, resulting kinetic changes of those parameters (in %) were averaged among volunteers with respect to timepoint of measurement and OCA mixture applied. This was done to make the observed changes more consistent as the initial values (intact skin measurements) are different between the volunteers due to interpatient variability.
3 RESULTS AND DISCUSSION
Scanning the manually established fit limits did not significantly affect the estimation of skin optical properties. The scattering coefficient µ s(λ 800) mean standard deviation (SD) for all OCA and for all timepoints was only layer, respectively. The same values for the scattering anisotropy factor g (λ 800) were ~0.05% and ~5%. Thus, there was almost no influence of manual fitting range delineation on estimated optical properties.
Figure 3 shows the average of two LC-OCT intensity profilesR (z ) before and after OC, together with the corresponding linear regression fits of the epidermal and dermal layers with the corresponding µ s values. It can be seen from the round insets and the corresponding pixel intensity distributions that the image contrast and brightness from dermal layer (at 200 µm depth) is increased after optical clearing.