Automated screening of Papanicolaou tests: A review of the literature

Abstract

Automated Papanicolaou test screening systems have now been available for over 25 years. Currently two automated screening systems are in widespread clinical use. These are the ThinPrep Imaging System and the FocalPoint GS Imaging System. In their current configurations, both facilitate faster screening by showing a limited number of fields of view (FOV) to cytotechnologists. The FOV are based on the use of proprietary algorithms applied to computerized images of the slide that determine the cells and cell groups with the highest likelihood of abnormality. If all of the FOV are deemed to be negative, the case can be signed out with no additional review; if one or more fields appear possibly abnormal, the entire slide must be manually screened. The United States Food and Drug Administration has ruled that for workload calculation purposes, looking at only the FOV review counts as one-half slide, potentially greatly increasing the number of slides that can be screened. However, follow-up studies of this technology have shown that screening accuracy declines when very large numbers of cases are reviewed per day. Recommendations designed to limit screening volumes to levels that do not jeopardize patient care have therefore been created. The development of fully automated screening that does not rely on human judgment remains an unrealized aspiration. This review covers the history of the development and clinical implementation of automated screening technology with descriptions of the various automated screening systems and their performance as reported in published literature.

Impact of image analysis and artificial intelligence in thyroid pathology, with particular reference to cytological aspects

“…To date, numerous computer algorithms for a variety of haematoxylin‐eosin‐stained histopathology slides have been developed to aid pathologists in rendering a reproducible and accurate diagnosis . Similarly, computer‐assisted screening systems have been effectively utilized to analyse Papanicolaou (Pap) smears for several decades . More recently, successful attempts to apply deep learning algorithms to WSI of Pap test slides have delivered promising results despite technical difficulties experienced with scanning cytology material …”

CytoProcessor^TM: A New Cervical Cancer Screening System for Remote Diagnosis

“…In spite of these difficulties, two automated screening systems have reached the market and attained considerable success [3]. These automated screening systems are direct descendants of technology from the 1980s, namely, AutoPap, now called the BD FocalPoint Slide Profiler (Becton, Dickinson and Company, Franklin Lakes, NJ, USA), and PAPNET, renamed as the ThinPrep Imaging System (TIS; HOLOGIC, Marlborough, MA, USA) [2].…”

“…In the case of the TIS, the screener places the slide on the microscope, and the system then positions the slide on the first selected field of view [3]. Once examined, the screener passes to the next field of view, until all 22 have been interpreted.…”

“…Current automated screening systems select and present a fixed number of “fields of view” for each slide [3]. The same number of fields of view are selected whether the slide contains zero abnormal cells or hundreds of abnormal cells.…”

Adaptation of CytoProcessor for cervical cancer screening of challenging slides

“…4 Both of these systems identify abnormal cells based on nuclear staining characteristics, imposing a dedicated sample processor and a fixed staining protocol. 4,5 One of the major costs for a laboratory in the implementation of these systems is the replacement of the existing sample preparation equipment. 6 CytoProcessor (DATEXIM, Caen, France) was designed to integrate into the laboratory's existing environment, without requiring the replacement of equipment.…”