Su1965 Digital Microscopy Is a Valid Alternative to Conventional Microscopy for Diagnosing Barrett's Esophagus

Abstract

Introduction Management of Barrett esophagus (BE) relies heavily on the histopathological assessment of biopsies. This assessment is subjective and associated with significant intraand inter-observer variation. Most guidelines recommend review of biopsies by expert pathologists in case of low-grade or high-grade dysplasia (LGD/HGD). Conventional review of microscopy slides, however, is impractical and does not allow intercollegiate conferences or annotations of relevant findings in images for feedback. A digital revision platform would overcome these practical limitations. Aims & Methods In preparation of a national BE digital revision platform we compared the diagnostic accuracy of conventional and digital microscopy for diagnosing BE+/-dysplasia by a panel of expert pathologists. Sixty BE biopsy slides (NDBE; n=25, LGD; n=20; HGD; n=15) were scanned at x20 magnification. Five expert BE pathologists independently assessed all slides four times in 2 alternating rounds of digital and conventional microscopy. Assessments were supervised by a research fellow and the order of rounds as well as the order of the slides was randomized. Pathologists were blinded for the original diagnosis and identifying slide features. Intra-observer and pairwise inter-observer agreement were calculated using custom weighted Cohen's kappa in four categories (NDBE; IND; LGD; HGD). Kappa scores were expressed as fraction of maximum possible kappa score for each cross table. Results The mean intra-observer agreement was 0.63 for the 2 rounds of digital assessment and 0.74 for the two rounds of conventional assessment. Mean pairwise inter-observer agreement was 0.61 and 0.64 for first and second round of digital microscopy, respectively. For the two rounds of conventional microscopy, mean pairwise inter-observer agreement was 0.62 and 0.66 respectively. In 48/ 60 (80%) of digital and in 50/60 (83%) of conventional microscopy reviews a majority diagnosis was reached after the first reading. Conclusion Diagnostic agreement of digital microscopy is comparable to conventional microscopy in the setting of an expert pathology platform for BE histology. This study validates the use of digital histopathological assessment of BE biopsies and will be used as the underlying infrastructure for a nationwide, web-based BE revision platform in the near future. This will overcome many logistical and practical issues concerned with conventional histologic review by multiple pathologists.