Search in the Abstract Database

Search Abstracts 2016

* = Presenting author

P013 Automated image analysis in the diagnosis of microscopic colitis: validation and implications for diagnosis and research

P. J. H. Engel*1, 2, L. K. Munck2, 3, U. H. Engel4, S. Holck4, A.-M. Kanstrup-Fiehn1, M. Kristensson5

1Roskilde Hospital, Department of Pathology, Roskilde, Denmark, 2University of Copenhagen, Faculty of Health and Medical Sciences, Copenhagen, Denmark, 3Koege Sygehus, Department of Gastroenterology, Koege, Denmark, 4Copenhagen University Hospital, Department of Pathology, Hvidovre, Denmark, 5Visiopharm, Hoersholm, Denmark

Background

The diagnosis of microscopic colitis (MC) rests on a triad of clinical symptoms, a normal or near-normal endoscopy and characteristic histopathological findings. Distinguishing between lymphocytic colitis (LC) and incomplete lymphocytic colitis (LCi) based on Haematoxylin-Eosin (HE) stained slides may be difficult. In borderline cases and in cases of uncertainty, it is therefore recommended to perform a CD3 staining to determine the precise number of intraepithelial T-lymphocytes (IELs). Automated image analysis (AIA) is useful for instance in research of diseases characterised by the presence of specific cell populations, eg, eosinophils in eosinophilic esophagitis, mast cells in Hodgkin’s lymphoma, and T- lymphocytes in lung allograft biopsies. This study aimed to develop software targeting intraepithelial T-lymphocytes (IELs) as an aid in discriminating between LC and LCi.

Methods

Software for AIA was developed with a training set of 10 colon biopsies (LC, LCi, and normal) to match manual scorings of IELs. The study set consisted of blinded biopsies from 59 patients previously diagnosed with LC or LCi. Four pathologists individually reviewed the biopsies and gave a diagnosis of LC or LCi. AIA was applied to count the number of T-lymphocytes in 3 compartments: Border (IELs of the surface epithelium), Crypts (IELs of the cryptal epithelium), and Tissue (T- lymphocytes of the whole biopsy). Diagnosis based on IEL counts obtained by AIA was compared with the consensus diagnosis provided by the 4 pathologists.

Results

The two-at-a-time, inter-pathologist diagnostic agreement was between 59% and 95%. The agreement between pathologists and AIA, was 97% (к = 0,858) in the border compartment 90% (к = 0,486) in the tissue compartment, and 67% (к = 0,323) in the cryptal compartment. The correlation between pathologists diagnosis and automated image analysis of IELs in the border compartment is shown in the table (NNF = normal or nonspecific findings). The agreement was 96.6% with a к = 0.858.

Table 1 Correlation between pathologists and AIA diagnosis

PathologistsDiagnosis inThe borderCompartment
NNFLCiLCtotal
AutomatedNNF10.0.1
Image analysisLCi0.628
BorderLC0.0.5050
CompartmentTotal165259

Conclusion

Diagnosing LC and LCi by counting CD3 stained lymphocytes by means of AIA is congruent with the diagnosis of trained pathologists. AIA may be useful as a supplementary diagnostic tool in borderline cases of LC and in differentiating between LC and LCi. Being an objective and reproducible tool, AIA eliminates inter-observer variability and could be tool to achieve uniform pathological diagnosis in multicentre studies. If applied in prospective cohort studies AIA could help to redefine the histopathological criteria of LC and LCi.