Automation

Home > Cytopathology > Automation

Cytopathology
Cytopathology techniques
Automation

Authors: Reid Wilkins, M.D., Abdallah Flaifel, M.D., Tamar C. Brandler, M.D., M.S.

Editorial Board Members: Bonnie Choy, M.D., Marc Pusztaszeri, M.D.

Last author update: 31 October 2022

Last staff update: 17 January 2023

Copyright: 2022-2024, PathologyOutlines.com, Inc.

PubMed Search: Cytopathology automation

Page views in 2023: 1,042

Page views in 2024 to date: 290

Table of Contents

Cite this page: Wilkins R, Flaifel A, Brandler TC. Automation. PathologyOutlines.com website. https://www.pathologyoutlines.com/topic/cytopathologyautomation.html. Accessed April 25th, 2024.

Definition / general

Automation in cytopathology refers to the process of slide preparation (fixation and staining), image acquisition and image analysis with identification of abnormalities by automated machinery in conjunction with cytologist review
Currently primarily utilized in gynecologic cytopathology

Essential features

Automation has mainly been implemented in gynecologic cytopathology with the development of automation systems that utilize liquid based cytologic processing
Automated screening has comparable sensitivity to manual screening with the added benefit of increased productivity
Increased utilization of whole slide imaging and artificial intelligence pose additional benefits and challenges to the development of a completely autonomous digital workflow in gynecologic screening

CPT coding

Gynecological
- 88112 - enriched / concentrated preparation (ThinPrep, SurePath)
- Automated screening:
  - 88175 - technical
  - 88141 - professional

Terminology

Analytic and quantitative cytology
Cytology image analysis techniques
Automated screening systems in cytopathology

Overview

Specimen collection
- Spatula / brush
Specimen processing
- Liquid based
  - ThinPrep
    - Methanol based PreservCyt fixative solution
    - Filtration and dispersion separate debris and mucus without adverse effect on cell appearance
    - Controlled pressure deposits cell layer in 20 mm diameter circle
  - SurePath
    - Preservative fluid (ethanol / methanol / isopropanol)
    - Centrifugation separates cells of interest
    - Cell layer placed in 13 mm diameter circle via sedimentation
Image acquisition
- Prepared slides loaded into imaging station
- Cell spot image scanned and sent to image processor
Image analysis (ACM Computing Surveys 2022;54:1)
- Segmentation approach
  - Thresholding, region, contour, texture, graph, clustering, deep learning
- Segmentation free
  - Support vector machine (SVM), fuzzy c-means (FCM), convolutional neural network
Cytotechnologist review
- Field of view (FOV) presented
- Slides without abnormality diagnosed as NILM and signed out
- Slide with abnormality marked for cytopathology review
Slide rescreening
- Entire slide manually rescreened and signed out

Advantages of automation

Increased productivity and cytotechnologist satisfaction
Greater efficiency in evaluating slides
Decreased fatigue and turnaround time
Lower hospital cost (deriving from increased productivity)
Reference: Diagn Cytopathol 2021;49:559

Disadvantages of automation

Similar sensitivity to manual screening (Cytojournal 2007;4:6)
Initial cost and maintenance
Additional quality assurance and quality control (continued calibration and validation of machines)
Requires additional training of cytopathologists and cytotechnologists
Remaining fatigue, inattention and habituation with increased number of slides necessitates implementation of daily slide limitation
Reference: Diagn Cytopathol 2021;49:559

Implementation

Gynecologic
- PAPNET (1994, Neuromedical Systems, Inc., Suffern, NY)
  - First truly digital pathology device
  - Neural network technology used to identify abnormal cells in cases already screened as negative by cytotechnologist
  - 128 abnormal cells / clusters identified, photographed and mailed back to laboratory as a digital tape for review by cytotechnologist
  - Additional resources: Laboratory Medicine 1991;22:276
- AutoPap 300QC (1995, NeoPath Inc., precursor to BD FocalPoint GS imaging system)
  - Originally developed for quality control of screened Pap smear slides to reveal false negatives
  - Cell annotations used to develop algorithms to score overall slides and individual field of view (higher score, more abnormalities)
  - Higher cumulative score, increased likelihood of finding abnormality on slide
  - Additional resources: Acta Cytol 1996;40:45
- ThinPrep imaging system (2004, HOLOGIC, Marlborough, MA)
  - FDA approved
  - ThinPrep specimen processing
  - ThinPrep image processer consists of imaging station, image processor controller, server and user interface
    - Imaging station: scans entire slides
    - Imaging processor controller: creates images and analyzes data
    - Server: stores slide and imaging data
    - User interface: allows operator to use the machine
  - Image processor selects 22 fields of view (FOV) at 100x magnification for review by cytotechnologist
  - Additional resources: HOLOGIC: ThinPrep Imaging System [Accessed 9 August 2022]
- BD FocalPoint GS imaging system (2001, Becton, Dickinson and Company, Franklin Lakes, NJ)
  - FDA approved
  - SurePath specimen processing
  - Image processor utilizes slide profiler and review station
    - Slide profiler screens slides and locations based upon likelihood of containing abnormality
    - Review station arranges slides into quintiles based upon likelihood of abnormality (1 = highest risk, 5 = lowest risk)
      - 10 FOVs for each slide are presented
      - Lowest quintile of slides (least likely to have abnormality) can be archived without further review
  - Additional resources: BD: FocalPoint GS Imaging System [Accessed 9 August 2022]
- Cytoprocessor (2017, DATEXIM, Caen, France)
  - CE certified (approval in Europe)
  - Utilizes any preparation protocol and liquid based cytology (LBC)
  - Detects cells in whole slide images and arranges depending on abnormality
  - Can be integrated into a variety of workflows without requiring additional materials
  - Additional resources: DATEXIM: CYTOPROCESSOR [Accessed 9 August 2022]

Effectiveness

MAVARIC trial (Health Technol Assess 2011;15:1)
- Conducted in England, compared Imager / FocalPoint to manual screening for potential implementation in National Health Service (NHS)
- Significantly lower sensitivity of automated screening detection of cervical intraepithelial neoplasia grade II (CIN 2) and above (0.92) compared to manual screening
- Comparable rates of detecting CIN 2 and above, between ThinPrep and SurePath
- Increased productivity in automated screening arm (60 - 80% higher)
Additional studies in Australia, Scotland and U.S.
- Findings:
  - Significantly lower inadequate / negative reporting rates, higher low grade reporting rates
  - No significant difference in detection of high grade squamous intraepithelial lesion (HSIL)
  - Increased specificity with imager assisted screening compared to manual screening alone (Cytopathology 2013;24:235)
  - No significant difference in positive predictive value of Imager screening versus manual screening of ThinPrep slides
  - Up to 27% increase in productivity using Imager screening versus manual screening of ThinPrep slides (Diagn Cytopathol 2007;35:96)

Implication

Daily screening limitations (Diagn Cytopathol 2019;47:20)
- Automated slides without abnormality count as 0.5 slides
- Abnormal automated screens requiring rescreening count as 1.5 slides
- Limits vary by country
  - U.S.: 100 slides/day
  - U.K.: 32 slides/day
  - Italy: 25 - 50 slides/day
  - Australia: 70 slides/day
  - Canada: 80 slides/day
- American Society of Cytopathology guidelines (Diagn Cytopathol 2013;41:174)
  - Limit Pap screening to < 7 hours/day
  - Average < 70 slides/day
- Computational cytology
  - Combines whole slide imaging with artificial intelligence
    - Advantages:
      - Potential increase in accuracy of slide interpretation
      - Further increase in productivity
    - Limitations (Acta Cytol 2021;65:301):
      - Requires established digital workflow
      - Regular calibration and maintenance of whole slide imaging and software packaging
      - Cytologic processing artifact present in whole slide imaging impairs digital datasets used for whole slide analysis
        
        3 dimensional clustering necessitates image capture on multiple focal planes (z stacking)
      - Limited datasets for training
      - Added responsibility for cytologist