PPubMed17 Mar 2021Cited 123×

Point-of-Care Digital Cytology With Artificial Intelligence for Cervical Cancer Screening in a Resource-Limited Setting Digital Cytology for Cervical Cancer Screening in Resource-Limited Setting Digital Cytology for Cervical Cancer Screening in Resource-Limited Setting

Oscar Holmström, Nina Linder, Harrison Kaingu, Ngali Mbuuko, Jumaa Mbete, Felix Kinyua, Sara Törnquist, Martin Muinde, Leena Krogerus, Mikael Lundin, Vinod Diwan, Johan Lundin

Abstract

'Jumaa Mbete' 'Felix Kinyua' 'Sara Törnquist' 'Martin Muinde' 'Leena Krogerus' 'Mikael Lundin' 'Vinod Diwan' 'Johan Lundin'] This diagnostic study uses data from a single health center in Kenya to investigate the use of digital microscopy and artificial intelligence in a resource-limited area to detect abnormal cells in Papanicolaou tests.

A figure from Point-of-Care Digital Cytology With Artificial Intelligence for Cervical Cancer Screening in a Resource-Limited Setting Digital Cytology for Cervical Cancer Screening in Resource-Limited Setting Digital Cytology for Cervical Cancer Screening in Resource-Limited Setting — fig. from the paper

§ The Valyu brief

The paper, in five minutes

Read from the full text via Valyu Contents

A point‑of‑care AI‑based digital cytology system accurately detected cervical cell atypia in a Kenyan clinic, achieving AUC up to 0.97 and sensitivity up to 100%.

What’s new

i.Developed and implemented a portable point‑of‑care digital microscopy system with mobile‑network slide upload.
ii.Created and trained a deep‑learning model to detect cervical cell atypia on whole‑slide images.
iii.Validated the system in a Kenyan clinic, showing high diagnostic accuracy (AUC 0.94‑0.97, sensitivity up to 100%).
iv.Demonstrated that remote AI analysis can reduce pathologist workload and enable screening in resource‑limited settings.

How they did it

Pap smears were collected from 740 HIV‑positive women, digitized on a portable 18‑MP whole‑slide scanner, and uploaded via 3G/4G to a cloud platform. A deep‑learning system was trained on annotated slides and evaluated on a validation set of 361 adequate slides, comparing its results to expert pathologist assessments of both digital and glass slides.

What they found

General atypia detection: AUC 0.94, sensitivity 95.7% (95% CI 85.5‑99.5), specificity 84.7% (95% CI 80.2‑88.5) on 361 slides.
High‑grade atypia detection: AUC 0.97, sensitivity 85.7% (95% CI 67.3‑96.0), specificity 98.5% (95% CI 96.5‑99.5).
Compared with physical slide assessment, DLS sensitivity 100% for general and high‑grade atypia, specificity 78.4% and 93.3% respectively.
Negative predictive values were ≥96% for all atypia categories.

Caveats

✕The reference standard was expert cytology, not histologically confirmed biopsies.
✕The prevalence of significant atypia was low, limiting the number of positive cases.
✕The study was conducted at a single rural clinic, so findings may not generalize to other settings.
✕Only HIV‑positive women were included, which may affect applicability to broader populations.

§ The Valyu brief

Reading the full paper and taking notes. This takes a few seconds…

§ Ask this paper

Ask a question about this paper

Valyu reads the full text and answers from what the paper actually says.

§ Related work in other archives

Point-of-Care Digital Cytology With Artificial Intelligence for Cervical Cancer Screening in a Resource-Limited Setting Digital Cytology for Cervical Cancer Screening in Resource-Limited Setting Digital Cytology for Cervical Cancer Screening in Resource-Limited Setting

The paper, in five minutes

Ask a question about this paper

Similar papers in all four archives

Similar papers in all four archives

The paper, in five minutes