[IPVS 2026] Improving Diagnostic Sensitivity in Cervical Cytology: A Comparative Study of an Automated AI-Integrated Workflow versus Manual Microscopy

Abstract

This study evaluated the impact of miLab™ CER, an all-in-one digital cytology platform integrating automated staining, digital imaging, and on-device AI analysis, on inter-observer agreement in cervical cytology, compared with conventional light microscopy. A total of 150 liquid-based cytology (LBC) slides, balanced across The Bethesda System (TBS) diagnostic categories, were independently reviewed by three cytologists under both conditions with a 2-week washout period. With miLab™ CER, inter-observer agreement improved from moderate to almost perfect (Fleiss’ κ 0.553 → 0.810), and overall reader agreement increased from 47.4% to 60.9%. Agreement on clinically critical HSIL cases rose from 0% to 27.8%. The results show the potential of miLab™ CER to deliver standardised, reproducible, reader-independent cervical cytology screening.

Table 1: Inter-Observer Agreement: miLab™ CER vs. Microscopy

Keywords