Machine learning to identify non-invasive mechanical testing approaches to predict skin composition, micro-structure and progression of ageing

This project aims to use machine learning algorithms to establish a predictive link between the structure and composition of the skin and non-invasive mechanical tests. Healthy male volunteers aged 20-30, 45-55 and 70-80 will be recruited from White Caucasian backgrounds. Photo-protected and photo-exposed skin sites will be characterised.
Non-invasive in vivo measurements will be made for all volunteers (mechanical and friction response, surface geometry, real-time imaging to characterise tissue deformation under loading and ATR-FTIR to characterise skin surface characteristics [lipids and moisture]).
Biopsy collection and testing will be undertaken from a sub population of volunteers from each age group. Skin protein composition and damage (biopsy 1) will be assessed by conventional mass spectrometry and peptide location fingerprinting. Skin mechanical and interaction response (biopsy 2) will be assessed using colloidal-probe micro-tribometry ([GTR] real-time force, in-situ microscopy under indentation and shear loading), surface characteristics (microgeometry using confocal microscopy [Olympus], surface lipids and moisture using FTIR [Perkin Elmer]) and structural characteristics (epidermal and dermal thickness, DEJ shape, primary collagen and elastin orientations).
Machine learning algorithms and multivariate (principal component) analysis of the combined data sets will aim to identify a predictive link between non-invasive measurements and skin structure/composition.