Clinical validation of primary diagnosis and feature detection by Hamamatsu NanoZoomer S360MD Digital Slide scanner system: Key Findings and Study Insights
In this presentation we will highlight two studies using the Hamamatsu S360MD Digital Slide scanner system (S360MD) were performed: 1) to compare the S360MD to standard conventional light microscopy (Glass) for primary diagnosis (PD); and 2) for precision of specific feature detection (FD) at different magnification levels. We will provide key findings from the study and insights.
Methods: Four sites participated in the 2 studies. In the PD study, each site was assigned 500 of 2,000 cases, across a range of organs, procedures, and diagnoses. Cases selected by an enrolling pathologist (EP) and a verifying pathologist (VEP) included the case’s ground truth primary diagnosis (GTPD), subtype, grade, margin, lymph node involvement, and a representative subset of slides. Four Reader Pathologists (RP) at each site diagnosed cases via both Glass and S360MD (randomly ordered), with a 4-week wash-out period between modalities. Adjudicator pathologists judged per pre-specified rules and a standardized harm table if each RP diagnosis “matched” the GTPD, or if a “major” or “minor” discrepancy existed. The primary endpoint was the difference between modalities in the major discordance rate. The upper bound of the 95% confidence interval (CI) of that difference had to be <4.0%.
In the FD study, a different EP and VEP enrolled 540 slides with 30 different primary histological features, each randomly placed within a field of view (FOV), with 10 FOVs each at 10x, 20x, and 40x. Coincidental secondary features were also identified. FOVs were pre-assigned to be “study” (n=378) or “sham” (n=162) cases. Three sub-studies were performed to assess the impact of scanning variables: intra-scanner (3 scans each on 3 scanners at 1 site), inter-scanner (1 scan each on 3 scanners at 1 site), and inter-site (1 scan each on 3 scanners, one at each of 3 sites). The scans were then read in random order by 9 different RPs at 3 sites who identified all observed features in each FOV from pre-specified list. The primary endpoint for each sub-study was the Average Positive Agreement (APA) between scans for each sub-study as follows: within a scanner for intra-scanner; between scanners for inter-scanner, and between sites for inter-site. The lower limit of the 95% CI for APA had to be >85%.
Results: In the PD study, the 2,000 cases and 5,705 slides resulted in 16,000 total reads. Five reads were deferred by RPs, three 360MD, two Glass. The study met the pre-specified acceptance criterion (see table below). In the FD study, despite 45% of FOVs having potentially confounding secondary features, each sub-study met the pre-specified acceptance criterion (see table below).
Conclusions: Primary diagnosis by 360MD was non-inferior to Glass with results comparable to previous studies. Feature detection by 360MD was at a very high rate, met study acceptance criteria and was comparable to a previous study. which will be helpful for attendees as they are designing their own studies or want to understand the regualtory landscape of the whole slide imaging devices. The presenters will highlight the challenges and barriers of such studies as well.
- To compare the S360MD to standard conventional light microscopy (Glass) for primary diagnosis (PD)
- Provide details of precision of specific feature detection (FD) at different magnification levels
- To provide key findings from the study and insights from such studies to help gain an understanding of the regulatory landscape of WSI devices
Anil Parwani, MD, PhD, MBA
Professor of Pathology
The Ohio State University
Dr. Anil Parwani is a Professor of Pathology at The Ohio State University. He serves as the Vice Chair and Director of Anatomical Pathology. Dr. Parwani is also the Director of Pathology Informatics and Director of the Digital Pathology. His research is focused on diagnostic and prognostic markers in bladder, prostate and renal cell carcinoma. Dr. Parwani has expertise in the area of surgical pathology, viral vaccines and pathology informatics including biobanking, whole slide imaging, digital imaging, telepathology, image analysis, artificial intelligence and lab automation. Dr. Parwani has authored over 375 peer-reviewed articles in major scientific journals and several books and book chapters. Dr. Parwani is the Editor-in-chief of Diagnostic Pathology and Journal of Pathology Informatics.
Liron Pantanowitz, MD, PhD, MHA
Chair, Department of Pathology
Professor Liron Pantanowitz is Chair and Professor of Pathology at the University of Pittsburgh in the USA. He received his medical degree and PhD from the University of Witwatersrand in South Africa. He completed his anatomical and clinical pathology residency training at Beth Israel Deaconess Medical Center, Harvard in Boston. He subsequently completed a hematopathology fellowship at Harvard and Cytopathology fellowship at Tufts. He is also board certified by the American Board of Pathology in clinical informatics and completed his MHA at Ohio University. Dr. Pantanowitz is an Editor-in-Chief of the Journal of Pathology Informatics. He is the president of the Digital Pathology Association, president of the American Society of Cytopathology, and a past president and current council member of the Association of Pathology Informatics. He is widely published in the field of pathology informatics and cytopathology. His research interests include digital pathology and artificial intelligence.