APIII - Advancing Practice, Instruction & Innovation Through Informatics

Whole Slide Image Based Interpretation of Immunohistochemistry Stains in Challenging Prostate Needle Biopsies

Jeffrey L Fine MD1 (finejl@upmc.edu) ; Jonhan Ho MD1; Yukako Yagi1; Drazen Jukic MD PhD1,2; John R Gilbertson MD3; Sheldon I. Bastacky MD2; Dana M. Grzybicki MD PhD1; Leslie Anthony1; Robb Wilson1; Anil V. Parwani MD PhD1,2; 1Centers for Pathology and Oncology Informatics, University of Pittsburgh; 2 Department of Pathology, University of Pittsburgh Medical Center; 3 Department of Pathology, Case Western Reserve University School of Medicine

Context:  Several hospitals in our academic network are supported by a centralized immunohistochemistry (IHC) laboratory, necessitating physical transportation of IHC stained slides.  Electronic distribution of these slides, using whole slide image (WSI) technology, could decrease turnaround time and courier-associated costs.  This study is a validation study of IHC stains in challenging prostate needle biopsies, retrospectively comparing glass microscope slides to WSI.

Technology:  WSI, also called “virtual slides”, were created using a high-throughput robotic scanner utilizing 20x objective magnification at a spatial sampling period of 0.47 microns per pixel (T2, Aperio Technologies, Vista, California, USA).  A server provided storage and access to WSI (Windows Server 2000, Microsoft, Redmond, Washington, USA).  Images were viewed using standard desktop computers using either a web  browser plug-in or stand-alone software, both provided by the imaging robot vendor.

Design:  One hundred consecutive prostate needle biopsy cases were screened by a non-participant who selected 30 cases with dotted foci which had been evaluated by IHC.  The IHC slides were scanned using a T2 scanner (Aperio Technologies, Vista, California, USA) and viewed using ordinary desktop computers.  Data was collected from 5 pathologists in 3 phases: 1) review of glass H&E slides; 2) review of IHC stains using WSI; 3) review of all original glass slides and 4) comparison of study findings with original diagnoses.

Results:  Preliminary analysis of phase 1 and 2 data showed high inter-observer variability for both phase 1 (agreement in 7 foci) and phase 2 (agreement in 15 foci).  For phase 1 (H&E only) this is expected.  However for phase 2 this high variability is not expected and could represent a true difference between WSI and glass slides.  Analysis of phase 3 data will be required to explore this further.  Approximately 79% of foci required fewer than 15 minutes to interpret.  Average confidence, rated from 1 (low) to 3 (high), was similar for phase 2 (2.6) and for phase 1 (2.4).

Discussion:  WSI validation studies serve to: highlight deficiencies in existing image quality and virtual microscopy systems; spur developers to address shortcomings; and provide pathologists with exposure to and knowledge of these emerging technologies.  Electronic IHC stain distribution, if feasible, would decrease temporal and financial costs associated with obtaining IHC from a central laboratory.