Introduction Fourier transform infrared spectroscopy (FTIR) detects specific molecules in human tissue from unique vibrational absorption spectra in the mid Infra-Red region (1800–900 cm^-1). Single element Attenuated total reflectance (ATR) contains averaged FTIR spectra from the superficial 10 microns of an entire oesophageal biopsy. We aim to extract individual cellular characteristics with molecular resolution by using FTIR and ATR on ex-vivo oesophageal biopsy specimens from patients undergoing endoscopy for BE surveillance or neoplasia assessment to detect high grade dysplasia (HGD).

Methods 731 spectra of 374 fresh biopsies from 76 patients were analysed. Biopsies were taken from visible BE. Before being placed in formalin, they were analysed by a spectrometer fitted with liquid nitrogen-cooled detector and ATR silicon microprism. For each spectrum 500 interferograms were averaged before Fourier transformation. Spectra were pre-processed using MATLAB scripts by spectrally removing liquid water and water vapour contributions, vector normalising to the 1610–900 cm^-1 region and second derivative conversion to remove baseline artefacts. Specific cellular characteristics were first determined. Unstained 8 µm tissue sections from 1 patient were analysed with FPA (Focal Plane Array)-FTIR imaging and correlated with stained slides. It was possible to accurately describe specific features ofsquamous epithelium (SQ), columnar lined epithelium (CLE), and lamina propria (LP) with this method. These features were applied to the 374 fresh biopsies using ATR-FTIR. Combined clustering and partial least squares regression discrimination (PLSDA) was used to build a diagnostic pipeline. Biopsies were grouped according to their cellular characteristics from the prior FTIR imaging. (1. SQ vs Rest, 2. SQ only biopsies, 3. CLE only biopsies, 4. CLE and LP containing biopsies and 5. LP containing biopsies only).

Results We distinguished SQ mucosa from CLE (BE), HGD and OAC tissue at an overall sensitivity of 89% and specificity of 91%. By grouping the spectra into groups according to their cellular contents, HGD was distinguished from all other biopsies with sensitivities and specificities of 68 and 89% (CLE only), 74 and 82% (CLE and LP) and 94 and 97% (LP only) respectively.

Conclusion Combined FTIR and ATR-FTIR spectroscopy can accurately distinguish HGD arising in BE on ex-vivo biopsy specimens and might become accurate enough to exclude routine histopathological evaluation in these patients.

Disclosure of Interest None Declared.