Article Text


PTU-168 Barrett’s Oesophagus Screening: Infrared Spectroscopy For Cytological Assessment
  1. OJ Old1,
  2. M Almond1,
  3. G Lloyd1,
  4. D Townsend2,
  5. K Lenau2,
  6. M Diem2,
  7. H Barr1
  1. 1Gloucestershire Hospitals NHS Trust, Gloucester, UK
  2. 2Northeastern University, Boston, USA


Introduction Screening for Barrett’s oesophagus could allow early detection, enabling timely diagnosis and intervention for oesophageal adenocarcinoma. Recent studies have shown the acceptability of a swallowed cytology brush (‘Cytosponge’) for cell collection. If introduced, cytological assessment would pose several challenges. Firstly, oesophageal cytology is performed infrequently, and expertise in this field is correspondingly limited. Secondly, assessment of individual cells is challenging even for experienced cytopathologists, with a degree of interobserver variability. Thirdly, screening would require a great deal of cytopathology resources. Infrared spectroscopy (IR) gives reproducible spectra based on cell biochemistry; applying multivariate statistical analysis and computer modelling can provide robust and rapid discrimination between pathological cell subtypes. We aimed to demonstrate the potential application of IR in analysis of oesophageal cytology.

Methods Endoscopic cytology brushes were used to collect oesophageal cells from patients undergoing endoscopy for Barrett’s oesophagus. Cells were fixed in formalin, centrifuged and slides prepared. IR spectra were measured across the entire sample area. Pre-processing steps allowed spectra from individual cells to be reconstituted. Further pre-processing removed confounding effects and enhanced signal-to-noise ratios. Conventional cytology analysis was undertaken to provide a reference for developing a predictive model using IR data. Chemometric analysis was then undertaken using Partial Least Squares Discriminant Analysis (PLS1DA) and cross-validation performed.

Results 23 cytology brush samples were collected from 11 patients. 4 samples contained low cell counts and were excluded from analysis. 5536 cells (2339 normal squamous, 2511 Barrett’s oesophagus and 686 dysplastic) were used to create and validate a predictive model. The predictive capability of the model is shown in the table below:

Abstract PTU-168 Table 1

Conclusion The high accuracy demonstrated by our predictive model suggests IR is a promising candidate for cytological analysis of oesophageal cells. As an objective, automated system, this technique could prove invaluable for Barrett’s screening in future.

Reference Schubert JM et al. Lab Invest. 2010;90(7):1068–1077

Disclosure of Interest None Declared.

Statistics from

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.