The clinical application of multispectral fluorescence lifetime imaging of human skin using multiphoton microscopy

Abstract

The work presented in this thesis employed multiphoton microscopy of tissue autofluorescence to investigate spectrally and fluorescence lifetime resolved images obtained from normal skin and cutaneous malignancies. This was achieved by adapting a commercially available CE-marked multiphoton tomograph (DermaInspect®) to allow fluorescence lifetime imaging (FLIM) simultaneously in four spectral channels and corresponding steady-state hyperspectral images using a prism-based spectrometer to be acquired. The images generated were analysed through the manual identification of morphological criteria and through manual and automatic segmentation of individual cells within FLIM images followed by automated morphological and spectroscopic analysis. The analysis of FLIM images acquired from normal skin ex vivo and in vivo identified subpopulations of cells based on their autofluorescence characteristics and allowed intra- and interpatient variations to be assessed. The mean cellular lifetime was found to decrease between 691-1286 picoseconds (ps) with depth, increase between 199-550 ps with age and a statistically significant decrease between 286-1436 ps with skin phototype (I-IV) was found, depending on spectral channel. The manual identification of morphological features from BCC images acquired ex vivo allowed the correct diagnosis to be made with a sensitivity/specificity of 79%/93%. Cellular fluorescence lifetimes were statistically significantly longer by between 19.9-39.8% compared to normal skin. A linear discriminant analysis combining both spectroscopic and morphological cellular parameters allowed BCCs to be discriminated from normal skin with an AUC of 0.83. Manually identified morphological features were able to distinguish dysplastic naevi from melanomas with a sensitivity and specificity of 75% and 81% respectively from ex vivo FLIM images. However, no contrast in cellular fluorescence lifetime was observed. A motorised stage has also allowed multispectral FLIM image mosaics of depth resolved images from unsectioned skin to be presented for the first time. In conclusion tissue autofluorescence and FLIM detect clinically useful differences in the skin.