Automated image analysis in the study of lymphocyte subpopulation in eosinophilic oesophagitis

BACKGROUND: Eosinophilic oesophagitis (EoE) is characterized by the presence of eosinophils in oesophageal mucosa. Other inflammatory cells, mainly lymphocytes, dendritic cells, and mast cells may also play an important role in this disease. The aim of this study is to compare the inflammatory pattern of the mucosa between EoE and gastro-oesophageal reflux disease (GERD), using automatic image analysis in digital slides, and to assess treatment response after elimination diet and food challenge test. METHODS: From 2010 to 2013, 35 oesophageal biopsies from EoE and GERD patients were randomly selected. In six EoE biopsies, patients had been treated with selective food exclusion diet. Immunohistochemical study with CD3, CD20, CD4, and CD8 for lymphocyte populations, CD1a for dendritic cells, and CD117/c-kit for mast cells was performed. Slides were scanned using Leica Aperio Scanscope XT with 40× magnification. Immunohistochemical expression was quantified in 245 immunohistochemistry digital slides with Leica Aperio positive pixel count algorithm using two different approaches: whole slide analysis versus selection of a 2 mm(2 )hot spot area. RESULTS: Average eosinophil cell count was significantly higher (p < 0.001) in the first biopsy of EoE patients before treatment (30.75 eosinophils per high power field - HPF) than in GERD patients (0.85 eosinophils/HPF) or in EoE patients after treatment with elimination diet (1.60 eosinophils/HPF). In the immunohistochemical study, manual count and automatic image analysis showed a significant increase in the number of CD3 and CD8 cells in EoE patients, compared with GERD patients. However, the increase of CD117/c-kit was only statistically significant when manual counting procedures were used. CD20 positive cell count also showed a non-statistically significant tendency to reduce after elimination diet treatment. Manual eosinophil count correlated much better with CD3 and CD8 count using hot spot approach than with a whole slide approach. CONCLUSIONS: Positive pixel count algorithm can be a useful tool to quantify the immunohistochemical expression of inflammatory cells in the diagnosis and follow up of eosinophilic oesophagitis.