Supplementary Material for: A Multi-Institutional Feasibility Study on the Use of Automated Screening Systems for Quality Control Rescreening of Cervical Cytology
Datasets usually provide raw data for analysis. This raw data often comes in spreadsheet form, but can be any collection of data, on which analysis can be performed.
Objective: To evaluate the efficacy of the automated screening system FocalPoint for cervical cytology quality control (QC) rescreening. False-negative rates (FNRs) were evaluated by a multi-institutional retrospective study. Study Design: Cervical cytology slides that had already been reported as negative for intraepithelial lesion or malignancy (NILM) were chosen arbitrarily for FocalPoint rescreening. Slides stratified into the highest 15% probability of being abnormal were rescreened by a cytotechnologist. The slides that were abnormal were reevaluated by a cytopathologist to be false negatives. Results: Rescreening of 12,000 slides, i.e. 9,000 conventional slides and 3,000 liquid-based cytology (LBC) slides, was performed; 9,826 (7,393 conventional and 2,433 LBC) were satisfactory for FocalPoint (2,174 were determined unsatisfactory) and those within the highest 15% of probability (1,496, i.e. 1,123 conventional and 373 LBC) were rescreened. As a result, 117 (96 conventional and 21 LBC) were determined as abnormal (other than NILM) and the FNR was 1.19%. Among these 117 slides, 40 (35 conventional and 5 LBC) were determined as high-grade squamous intraepithelial lesion and greater (HSIL+). Conclusion: Of 117 (1.19%) abnormal slides detected, 40 (0.41%) were determined to be HSIL+. This result suggests that FocalPoint is effective for QC rescreening of cervical cytology.