Article Text

Download PDFPDF

PTU-143 Digital Image Analysis Enhances Quantitative Immunohistochemistry in the Squamous-Metaplasia-Dysplasia-Carcinoma Sequence
  1. E S Bloom1,
  2. M A Butt1,2,
  3. R J Haidry1,2,
  4. D Oukrif3,
  5. S-U-R Khan1,
  6. S Ahammed3,
  7. V Sehgal1,2,
  8. J Louis-Auguste1,
  9. M R Banks1,2,
  10. M Gandy4,
  11. M Rodriguez-Justo3,
  12. M Novelli3,
  13. L B Lovat1,2
  1. 1NMLC, UCL
  2. 2GI services
  3. 3Pathology, UCH
  4. 4UCL-AD, UCL, London, UK


Introduction We have previously shown how the Allred scoring system may be used to semi-quantify expression of nuclear biomarkers in the Barrett’s (BE) to oesophageal adenocarcinoma (OA) sequence. Recently, a number of digital image analysis (DIA) platforms have been clinically validated for quantification of immunohistochemistry (IHC) in breast tissue. This study aims to compare pathologists scores (PS) with DIA for the quantification of nuclear biomarkers in BE to OA sequence.

Methods Paraffin embedded specimens were selected from 34 patients. Pathology grade (PG) was scored as 1 (non-dysplastic BE; n = 5), 2 (low grade dysplasia; n = 5), 3 (high grade dysplasia; n = 11) and 4 (OA; n = 14). Sections were immunostained with antibodies PLK1-M, PLK1-L and Geminin. Intensity (I-PS) (0 to 3+) and extent (E-PS) (0; < 1% = 1; 1–10% = 2; 10–33% = 3; 33–66% = 4; > 66% = 5) of staining were scored by 2 GI pathologists, and mean PS calculated. Intensity and proportions of +ve staining were digitally quantified using Ariol® software. Analysis classifiers were trained to identify thresholds of positive (brown/DAB) and negative (blue/Hx) nuclei (Figure A) in the areas of interest. Background tissue was digitally excluded. Mean intensity (I-DIA) and mean counts for DIA positive (red) and negative (green) nuclei (Figure B) were quantified to calculate percentage positivity (E-DIA) and compared with I-PS, E-PS and total Allred score (A-PS) using Pearson correlation coefficient.

Results Significant correlation was seen between E-DIA and A-PS (r = 0.76, p = 0.006; r = 0.73, p = 0.008; r = 0.94, p = 0.0004) with all biomarkers (PLK1-M; PLK1-L; Geminin). PLK1-L showed additional correlation between DIA and PS for intensity (r = 0.985, p = 0.02) and extent (r = 0.95, p = 0.90). Geminin showed additional correlations between DIA and PS for extent (r = 0.99, p = 0.0008) and PG (r = 0.97, p = 0.03). Following training, Ariol® analysis took a mean of 4mins (Range 3–5) per tissue region highlighted.

Conclusion This study has demonstrated how DIA may be used to quantify expression of nuclear biomarkers. Significant correlation between DIA with Allred score was seen with all biomarkers, but only PLK1-L correlated with intensity and Geminin with PG. Background staining ignored by pathologists was found to be a confounder for DIA, particularly with PLK1-M. Nonetheless, DIA has great potential to enhance current grading and risk stratification systems for BE, and help select patients for targeted therapies dependent on biomarker expression.

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.