AI-Assisted Cervical Cancer Screening

TL;DR

This work targets the subjectivity of VIA for cervical cancer screening in LMICs by delivering an end-to-end smartphone-enabled AI system. It collects a real-world dataset from 1,430 women (9844 images) across 32 camps using a handheld protocol and trains a dual-image $ResNet$-18 classifier on pre- and post-VIA images to predict VIA outcomes, with a quality-controlled annotation framework and automated cervix cropping. The model achieves a balanced accuracy of $74.62\%$, sensitivity $75.58\%$, and specificity $73.67\%$ on a held-out test set, while inter-rater reliability among experts is moderate ($\alpha=0.49$), highlighting inherent label noise. The study demonstrates the feasibility of using readily available smartphones for VIA data capture and AI-assisted decision support, and it lays a foundation for larger-scale validation and expanded tasks such as real-time cervix landmarks and TZ-type classification to further improve screening reliability in resource-constrained settings.

Abstract

Visual Inspection with Acetic Acid (VIA) remains the most feasible cervical cancer screening test in resource-constrained settings of low- and middle-income countries (LMICs), which are often performed screening camps or primary/community health centers by nurses instead of the preferred but unavailable expert Gynecologist. To address the highly subjective nature of the test, various handheld devices integrating cameras or smartphones have been recently explored to capture cervical images during VIA and aid decision-making via telemedicine or AI models. Most studies proposing AI models retrospectively use a relatively small number of already collected images from specific devices, digital cameras, or smartphones; the challenges and protocol for quality image acquisition during VIA in resource-constrained camp settings, challenges in getting gold standard, data imbalance, etc. are often overlooked. We present a novel approach and describe the end-to-end design process to build a robust smartphone-based AI-assisted system that does not require buying a separate integrated device: the proposed protocol for quality image acquisition in resource-constrained settings, dataset collected from 1,430 women during VIA performed by nurses in screening camps, preprocessing pipeline, and training and evaluation of a deep-learning-based classification model aimed to identify (pre)cancerous lesions. Our work shows that the readily available smartphones and a suitable protocol can capture the cervix images with the required details for the VIA test well; the deep-learning-based classification model provides promising results to assist nurses in VIA screening; and provides a direction for large-scale data collection and validation in resource-constrained settings.

Figures (2)

• Figure 1: Architecture of the model used(lower block) based on ResNet-18(upper block)
• Figure 2: Results: Confusion Matrix on the test set(Left), Visualization of sample images with false predictions(Right); in the Left of each pair is a pre-VIA image and in the right is a post-VIA image.