BOVNet: Cervical Cells Classifications Using a Custom-Based Neural Network with Autoencoders

Diogen Babuc, Darian Onchiş

2025

Abstract

Cervical cancer is a major global health challenge being the fourth-most common type of cancer. This emphasizes the need for accurate and efficient diagnostic tools that work well for small clinical datasets. This paper introduces an approach to computer-aided cervical scanning by integrating a custom-based neural network with autoencoders. The proposed architecture, Baby-On-Vision neural network (BOVNet), is tailored to extract intricate features from cervical images, while the autoencoders mitigate noise and enhance image quality. State-of-the-art architectures and the BOVNet architecture are trained on three comprehensive data sets (496, 484, and 1050 samples) that include Pap smear scans and histopathological findings. We demonstrate the effectiveness of our approach in accurately predicting cervical cancer risk and stratifying patients into appropriate risk categories. A comparative analysis with existing screening methods indicates the superior performance of BOVNet in terms of sensitivity (between 90.9% and 98.81% for three data sets), general predictive accuracy (between 92% and 94.86%), and time efficiency in identifying the increased risk of cervical abnormalities.

Paper Citation

in Harvard Style

Babuc D. and Onchiş D. (2025). BOVNet: Cervical Cells Classifications Using a Custom-Based Neural Network with Autoencoders. In Proceedings of the 11th International Conference on Information and Communication Technologies for Ageing Well and e-Health - Volume 1: ICT4AWE; ISBN 978-989-758-743-6, SciTePress, pages 173-180. DOI: 10.5220/0013201500003938

in Bibtex Style

@conference{ict4awe25,
author={Diogen Babuc and Darian Onchiş},
title={BOVNet: Cervical Cells Classifications Using a Custom-Based Neural Network with Autoencoders},
booktitle={Proceedings of the 11th International Conference on Information and Communication Technologies for Ageing Well and e-Health - Volume 1: ICT4AWE},
year={2025},
pages={173-180},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013201500003938},
isbn={978-989-758-743-6},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 11th International Conference on Information and Communication Technologies for Ageing Well and e-Health - Volume 1: ICT4AWE
TI - BOVNet: Cervical Cells Classifications Using a Custom-Based Neural Network with Autoencoders
SN - 978-989-758-743-6
AU - Babuc D.
AU - Onchiş D.
PY - 2025
SP - 173
EP - 180
DO - 10.5220/0013201500003938
PB - SciTePress