Automated Cervical Cancer Detection through Visual Inspection with Acetic Acid in Resource-Poor Settings with Lightweight Deep Learning Models Deployed on an Android Device

Leander Melroy Maben, Keerthana Prasad, Shyamala Guruvare, Vidya Kudva, P C Siddalingaswamy · 2025

Read Automated Cervical Cancer Detection through Visual Inspection with Acetic Acid in Resource-Poor Settings with Lightweight Deep Learning Models Deployed on an Android Device by Leander Melroy Maben, Keerthana Prasad, Shyamala Guruvare, Vidya Kudva, P C Siddalingaswamy (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-13253.txt.

Introduction

Leander Maben and co-authors automate cervical screening with light AI on a phone, in work from 2025. Cervical cancer kills many women in poor countries despite being treatable when caught early. A vinegar swab test is cheap but needs a trained eye, which makes readings vary.

Their system pairs a small object finder with a compact image network. The finder locates the cervix zone, and the network judges the swab result. The whole tool runs on a basic Android device with no cloud link. Task shifting to less trained health workers becomes feasible as a result. Screening camps could cover far more women with the same staff.

Catalog record

CREATORLeander Melroy Maben, Keerthana Prasad, Shyamala Guruvare, Vidya Kudva, P C Siddalingaswamy
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE35 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-13253

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Automated Cervical Cancer Detection through Visual Inspection with Acetic Acid in Resource-Poor Settings with Lightweight Deep Learning Models Deployed on an Android Device Leander Melroy Maben, Keerthana Prasad, Shyamala Guruvare, Vidya Kudva, P C Siddalingaswamy Abstract Cervical cancer is among the most commonly occurring cancer among women and claims a huge number of lives in low and middle-income countries despite being relatively easy to treat. Several studies have shown that public screening programs can bring down cervical cancer incidence and mortality rates significantly. While several screening tests are available, visual inspection with acetic acid (VIA) presents itself as the most viable option for low-resource settings due to the affordability and simplicity of performing the test. VIA requires a trained medical professional to interpret the test and is subjective in nature. Automating VIA using AI eliminates subjectivity and would allow shifting of the task to less trained health workers. Task shifting with AI would help further expedite screening programs in low-resource settings. In our work, we propose a lightweight deep learning algorithm that includes EfficientDet-Lite3 as the Region of Interest (ROI) detector and a MobileNet-V2 based model for classification. These models would be deployed on an android-based device that can operate remotely and provide almost instant results without the requirement of highly-trained medical professionals, labs, sophisticated infrastructure, or internet connectivity. The classification model gives an accuracy of 92.31%, a sensitivity of 98.24%, and a specificity of 88.37% on the test dataset and presents itself as a promising automated low-resource screening approach. [prefix=] [ orcid=0000-0002-4138-6814] [prefix=] [prefix=] [ orcid=0000-0002-7597-7591] keywords Automated Cervical Cancer Detection ,Android-Based Screening ,Low Resource Screening ,Artificial Intelligence ,Visual Inspection with Acetic Acid † † corresponding: Corresponding author 1 Introduction Cervical cancer is the fourth most common cancer among women globally. It caused an estimated 342,000 deaths in 2020 [ 1 ] and represents 6.5% of all female cancers [ 2 ] . The large number of deaths is of major concern, especially because cervical cancer is one of the most successfully treatable cancers when it is detected at an early stage and managed effectively. Cervical cancer develops in a woman’s cervix, which is the entrance to the uterus from the vagina [ 3 ] . Almost all cervical cancers are linked to infection with high-risk human papillomaviruses (HPV), an extremely common virus that transmits through sexual contact [ 2 ] . Persistent infection with high-risk HPV types may lead to the development of precursor lesions of the cervix, called CIN (cervical intraepithelial neoplasia), which represent epithelial cellular changes. CIN can be graded as CIN-1 (mild), CIN-2 (moderate), and CIN-3 (severe). Moderate and severe CINs are more likely to develop into invasive cancer, although a proportion of even these grades of CIN may regress or persist. Invasive cancer develops from CIN progressively from mild to moderate to severe and then cancer. This progression from pre-cancer to cancer takes 7 to 20 years [ 4 , 5 ] . The long lag between pre-cancer and cancer can be effectively used to screen, detect and treat pre-cancer before progression to cancer [ 6 ] . About 90% of new cervical cancer cases and deaths in 2020 occurred in low and medium-income countries [ 1 ] . Moreover, cervical cancer was found to be the most common type of cancer among women in 36 countries of the same income groups. Unsurprisingly, the low-resource countries which bore 90% of cervical cancer deaths had limited prevention, screening, and treatment services [ 2 ] . These staggering statistics highlight the need for a cost and resource-effective way to screen women for cervical cancer in highly rural and remote settings. Conventionally used screening practices for cervical cancer include cytology-based tests like Papanicolaou (Pap) smear or the newer liquid-based cytology technique, HPV tests, and visual inspection [ 7 ] . Soon after its discovery, the pap smear became the gold standard for cervical cancer screening and is still used in primary screening. Cells from the transformation zone are collected and transferred to a liquid preservative, followed by automated processing or inspection under the microscope for liquid-based cytology and conventional pap smear, respectively [ 7 , 8 ] . Pap smear and liquid-based cytology do not show considerable differences in sensitivity and specificity in detecting CIN. Pap smear has demonstrated consistent specificity of 98% but lower and variable sensitivity of 55%-80%. It relies on repeated screening throughout a woman’s lifetime to balance the low sensitivity. Cytology-based screening is not feasible for low-resource settings due to the requirement of electricity for microscopes, supplies for testing, trained cytopathologists, and repeated screening. This is coupled with the difficulty of patient follow-up in case of positive results and the poor sensitivity and specificity for high-grade lesions recorded in developing countries [ 7 ] . HPV tests can be used to detect HPVs. HPV tests also come with a self-swab variant where a patient can collect the vaginal sample by herself without the need for a medical professional. Testing for high-risk HPVs has been shown to achieve higher sensitivity than cytology-based tests in detecting high-grade CIN and can thus help prevent cervical cancer more effectively [ 9 ] . HPV testing is unsuitable for low-resource settings due to the high cost and the requirement of a laboratory. Visual Inspection with Acetic acid (VIA) is one of the techniques to screen for cervical cancer based on visual inspection. It involves visual inspection of the cervix after applying 3% to 5% acetic acid. Application of acetic acid at 3% to 5% causes reversible coagulation of cellular protein. Maximal coagulation occurs in areas with dysplasia or invasive cancer due to higher concentrations of proteins, making these areas appear acetowhite [ 4 ] . Examples of VIA-negative and VIA-positive cervix images are shown in figure 1 and figure 2 , respectively. VIA is low-cost, safe, and can be performed by a wide range of medical providers owing to its simplicity, making it suitable for use in low-resource settings. Results are provided instantly without the need for laboratory processing. Reviews have reported a sensitivity of 84% and specificity of 82% in detecting high-grade dysplasia [ 7 ] . Figure 1: Sample VIA-negative images. The figure shows cervixes that do not have aceto-white regions. Figure 2: Sample VIA-positive images [ 10 ] . The figure shows cervixes that have aceto-white regions. These regions have been marked with a dashed red circle. While VIA is a promising technique to curb cervical cancer mortality rates in remote areas, it still requires a trained professional to interpret the results. This poses a problem as the lack of professionals is often a major issue in resource-poor areas. Moreover, the accuracy of the test could vary based on the professional’s skills, as the test is subjective. Computer-based processing can go a long way in helping shift the task from highly trained professionals to less trained health workers and in neutralizing the variability in test performance as a result of variations in human skill. The aim of this study is to develop an algorithm that can process cervix images after the application of 3% to 5% acetic acid and detect the presence of pre-cancerous or cancerous lesions. Such an algorithm deployed on a portable android

Read the complete text