Engineering vascularized skin-mimetic phantom for non-invasive Raman spectroscopy

Abstract

Recent advances in Raman spectroscopy have shown great potential for non-invasive analyte sensing, but the lack of a standardized optical phantom for these measurements has hindered further progress. While many research groups have developed optical phantoms that mimic bulk optical absorption and scattering, these materials typically have strong Raman scattering, making it difficult to distinguish metabolite signals. As a result, solid tissue phantoms for spectroscopy have been limited to highly scattering tissues such as bones and calcifications, and metabolite sensing has been primarily performed using liquid tissue phantoms. To address this issue, we have developed a layered skin-mimetic phantom that can support metabolite sensing through Raman spectroscopy. Our approach incorporates millifluidic vasculature that mimics blood vessels to allow for diffusion akin to metabolite diffusion in the skin. Furthermore, our skin phantoms are mechanically mimetic, providing an ideal model for development of minimally invasive optical techniques. By providing a standardized platform for measuring metabolites, our approach has the potential to facilitate critical developments in spectroscopic techniques and improve our understanding of metabolite dynamics in vivo.

Figure 1.

Schematic illustration of fabrication method for skin tissue phantoms exhibiting a 3-D network of (a) open and (b) closed millifluidic channels.

Figure 2.

(a) Uniaxial compression test underway with MTS Criterion electromechanical system to determine mechanical properties of skin phantom; (b) Young’s modulus of skin phantom prepared with different concentrations of agarose performed in triplicates. The Young’s modulus of human epidermis and dermis have been highlighted; Normalized stress-relaxation curves with shaded error bar for the skin phantom with agarose concentrations of (c) 1%, (d) 2%, (e) 3%, (f) 4%, (g) 6%, and (h) 8% with three repeats for each concentration; (i) Stress-relaxation half-time for the skin phantom prepared with different concentrations of agarose performed in triplicates, with reported data for human skin highlighted.

Figure 3.

(a) Schematic illustration for fluorescence imaging of diffusion of 10 mg/mL 2-NBDG solution in the skin phantom. Fluorescence time-lapse images of fluorescent glucose (2-NBDG) molecules diffusing through (b) 2% agarose gel, (c) 2.5% agarose gel, and (d) 8% agarose gel.

Figure 4.

(a) Schematic for the Raman measurements performed on skin phantom; (b) Raman spectra of glucose solution at different concentrations; (c) Prediction versus true concentration after performing leave-one-sample-out partial least square (PLS) regression; (d) Coefficient of variation at each predicted concentration; (e) Variable importance in projections score for the PLS analysis showing that significant contributions come from the Raman peaks at 1064 cm⁻¹ and 1124 cm⁻¹.

Figure 5.

Photographs of dermis phantoms demonstrating designs achievable with the described method (a) straight channels; (b) curved channels; (c) 3-D channels; (d) 3-D branched networks; (e) open 3-D branched networks.