A Multicenter Study to Evaluate Diagnostic Accuracy by Pathologists Using the Aperio GT 450 DX in Local and Remote Viewing Stations.

Abstract

Context.—: The adoption of digital pathology may enable pathologists to perform primary diagnosis in both local and remote whole slide image viewing settings, improving logistics and convenience.

Objective.—: To test the performance of a new whole slide imaging system (Aperio GT 450 DX), both local intranet-based and remote internet-based viewing were compared with manual glass slide light microscopy.

Design.—: A total of 1161 curated cases, enriched with difficult clinical diagnoses, were enrolled in this accuracy study and digitally scanned on 3 Aperio GT 450 DX instruments at 3 clinical sites. Ten reading pathologists across the 3 study sites viewed images either locally (directly connected to the image server) or remotely (viewed over an internet connection). Each diagnosis was scored (concordant, minor, or major discrepancy) by a separate team of 3 adjudication pathologists. The diagnostic accuracy of the Aperio GT 450 DX was tested by comparing the whole slide image review diagnosis with the conventional light microscope manual slide review diagnosis.

Results.—: The difference in the major discrepancy rate between whole slide image review diagnosis and manual slide review diagnosis was 2.40% (95% CI, 1.40%-3.39%), meeting the predefined acceptance criterion of the 95% CI upper bound of 4% or less. Secondary end points were also met, including an upper bound of 7% or less and both local-only and remote-only upper-bound discrepancy rates of 4% or less. Major discrepancies were slightly lower for the remotely viewed cases (2.17%) compared with local direct server connection (2.61%), and time per read was not different.

Conclusions.—: The diagnoses made using the Aperio GT 450 DX, using both local and remote access image data, are noninferior to the diagnoses made using conventional light microscopy.