Search

Screening Papanicolaou test samples has proven to be highly effective in reducing cervical cancer-related mortality. However, the lack of trained cytopathologists hinders its widespread implementation in low-resource settings. Deep learning-based telecytology diagnosis emerges as an appealing alternative, but it requires the collection of large annotated training datasets, which is costly and time-consuming. In this paper, we demonstrate that the abundance of unlabeled images that can be extracted from Pap smear test whole slide images presents a fertile ground for self-supervised learning methods, yielding performance improvements relative to readily available pre-trained models for various downstream tasks. In particular, we propose \textbf{C}ervical \textbf{C}ell \textbf{C}opy-\textbf{P}asting ($\texttt{C}^{3}\texttt{P}$) as an effective augmentation method, which enables knowledge transfer from open-source and labeled single-cell datasets to unlabeled tiles. Not only does $\texttt{C}^{3}\texttt{P}$ outperforms naive transfer from single-cell images, but we also demonstrate its advantageous integration into multiple instance learning methods. Importantly, all our experiments are conducted on our introduced \textit{in-house} dataset comprising liquid-based cytology Pap smear images obtained using low-cost technologies. This aligns with our objective of leveraging deep learning-based telecytology for diagnosis in low-resource settings.

Additional file 1: Supplemental Table. Detailed breakdown of CHWs training costs for each session.

RENY_QuestionnaireFR (DOCX 118 kb)

Developing countries are interested in using human papillomavirus (HPV) testing as a primary screening test for cervical cancer prevention programs. The low specificity of the HPV assay requires triage testing of HPV-positive women. The aim of the study is to compare visual inspection with acetic acid (VIA) and cytology as triage testing methods in HPV-positive women to detect cervical intraepithelial neoplasia or Grade 2 or higher (CIN2+). The study was conducted in two Cameroonian towns (Yaoundé and Edea) and included 846 eligible women aged 25 to 65 years. All participants performed self-HPV testing. HPV-positive women (n = 259) were randomly assigned to be tested either by VIA (VIA group) or cytology (cytology group). HPV-positive women had both cervical biopsy and endocervical curettage to detect biopsy-confirmed CIN2+. All statistical tests were two-sided. The prevalence of HPV was 38.5%, and the mean age of HPV-positive women was 41.5 ± 10.1 years. One hundred ninety-eight women (97 in the VIA group and 99 in the cytology) were randomly assigned to one of the two testing arms. The sensitivity of VIA was 25.0% (95% CI, 7.1-59.1%), and the sensitivity of cytology was 90.0% (59.6-98.2%). The specificity was 74.2% (95% CI, 64.2-82.1%) for VIA and 85.2% (76.3-91.2%) for cytology. ROC area for cytology was 0.910 against the 0.496 area for VIA. In this trial, VIA was inferior to cytology as a triage test among HPV-positive women. Further investigations are needed to determine the optimal triage method for HPV-positive women.