Label-Free Quantification of Pharmacokinetics in Skin with Stimulated Raman Scattering Microscopy and Deep Learning

Abstract

The treatment of inflammatory skin conditions relies on a deep understanding of how drugs and tissue behave and interact. Although numerous methods have been developed that aim to follow and quantify topical drug pharmacokinetics, these tools can come with limitations, assumptions, and trade-offs that do not allow for real-time tracking of drug flow and flux on the cellular level in situ. We have developed a quantitative imaging toolkit that makes use of stimulated Raman scattering microscopy and deep learning-based computational image analysis to quantify the uptake of specific drug molecules in skin without the need for labels. Analysis powered by trained convolutional neural networks precisely identified features such as cells, cell junctions, and cell types within skin to enable multifactorial analysis of skin pharmacokinetics. We imaged and quantified the flow and flux of small molecule drugs through the layers and structures of exvivo nude mouse ear skin and extracted pharmacokinetic parameters through convolutional neural network-based image processing, including relative area under the curve accumulation, time of maximum drug concentration, and in situ partition ratios. This approach, which facilitates the direct observation and quantification of pharmacokinetics, can be used to glean mechanistic insight into underlying phenomena in skin pharmacokinetics.