Application of virtual microscopy in clinical cytopathology.


Journal Article

Virtual microscopy (VM) refers to the use of an automated microscope and digital imaging technology to scan, store, and view glass slides. VM systems allow the user to view a scanned image of the entire slide at multiple magnifications on a computer screen. We tested VM to evaluate its possible utility in diagnostic cytopathology. Ten cervical-vaginal monolayered preparations (AutoCyte preparation) were scanned using a BLISS (Bacus Laboratories Inc. Slide Scanner) system. Approximately 20-30% of the cellular area of each slide was imaged. The cases were randomly chosen to include examples ranging from benign cellular changes (BCC) to high-grade squamous intraepithelial lesions (HSIL). The computer performed image tiling and fusing of multiple JPEG images to create a high-quality VM slide. Six examiners (two each of cytopathologists, senior residents, and cytotechnologists) blindly evaluated the VM slides using an image server program (WebSlide Browser thin client software). The cytopathologic diagnoses made on the VM slide were then compared to the original glass slide diagnoses. BLISS took 36-100 min (avg. 58.4 min) to scan the selected fields in a glass slide with file sizes ranging from 23.1-83.6 MB. Time taken by the examiners to render a diagnosis ranged from 1-15 min (avg. 4.1 min) per case. The combined diagnostic accuracy was 98.3%. Only one case of LSIL was missed by one examiner. VM is a promising new tool, which gives a user the feel and simulated experience of an actual microscopic examination and provides a useful alternative to a glass slide in diagnostic cytopathology. Possible applications include: 1) second opinion consultation without transporting the glass slide, 2) education, 3) VM proficiency tests / board exams, and 4) telepathology. Shortcomings include 1) expensive initial setup, 2) inability to maintain an adequate focus in a thick smear with multiple levels, 3) large storage size of the VM slide, and 4) relatively long time needed to scan a slide.

Full Text

Duke Authors

Cited Authors

  • Steinberg, DM; Ali, SZ

Published Date

  • December 2001

Published In

Volume / Issue

  • 25 / 6

Start / End Page

  • 389 - 396

PubMed ID

  • 11747236

Pubmed Central ID

  • 11747236

International Standard Serial Number (ISSN)

  • 8755-1039

Digital Object Identifier (DOI)

  • 10.1002/dc.10021


  • eng

Conference Location

  • United States