MFEM-CIN: A Lightweight Architecture Combining CNN and Transformer for the Classification of Pre-Cancerous Lesions of the Cervix

Sep 21, 2026

Cervical cancer ranks among the most common cancers in women worldwide and remains a major cause of cancer-related deaths, especially in developing regions with limited medical resources and specialized physicians. Colposcopy-based detection supports early screening but often faces effectiveness constraints in underserved settings. This paper presents an automated colposcopic image analysis framework centered on MFEM-CIN, a hybrid CNN–Transformer model that aggregates local and global feature correlations for efficient clinical diagnosis. Experiments show 89.2% accuracy for identifying cervical intraepithelial neoplasia while maintaining a lightweight design, outperforming average physicians and supporting practical deployment for resource-constrained care.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

AccessFree to download

Abstract

Methods

Conclusions

I. INTRODUCTION

Cervical cancer burden

Screening tests

MFEM-CIN: A Lightweight Architecture Combining CNN and Transformer for the Classiﬁcation of Pre-Cancerous Lesions of the Cervix Peng Chen , Member, IEEE, Fobao Liu, Jun Zhang, and Bing Wang , Senior Member, IEEE Abstract—Goal: Cervical cancer is one of the most common cancers in women worldwide, ranking among the top four. Unfortunately, it is also the fourth leading cause of cancer-related deaths among women, particularly in developing countries where incidence and mortality rates are higher compared to developed nations. Colposcopy can aid in the early detection of cervical lesions, but its effectiveness is limited in areas with limited medical resources anda lack of specialized physicians. Consequently, many cases are diagnosed at later stages, putting patients at signiﬁcant risk. Methods: This paper proposes an automated colposcopic image analysis framework to address these challenges. The framework aims to reduce the labor costs associated with cervical precancer screening in undeserved regions and assist doctors in diagnosing patients. The core of the framework is the MFEM-CIN hybrid model, which combines Convolutional Neural Networks (CNN) and Transformer to aggregate the correlation between local and global features. This combined analysis of local and global information is scientiﬁcally useful in clinical diagnosis. In the model, MSFE and MSFF are utilized to extract and fuse Manuscript received 21 September 2023; revised 3 December 2023 and 6 January 2024; accepted 5 February 2024 . Date of publication 20 February 2024; date of current version 27 March 2024 . This work was supported in part by the National Natural Science Foundation of China under Grant 62072002, Grant 62172004, and Grant 62273001, in part by Anhui Province Collaborative Innovation Project under Grant GXXT-2022-050 and Grant GXXT-2022-053, in part by the Outstanding Research and Innovation Team Project of Anhui Province under Grant 2022AH010005, and in part by the Special Fund for Anhui Agriculture Research System under Grant 2021-2025 . The review of this article was arranged by Editor Amir Amini. (Corresponding authors: Jun Zhang; Bing Wang.) Peng Chen is with the National Engineering Research Center for Agro-Ecological Big Data Analysis and Application, Information Materials and Intelligent Sensing Laboratory of Anhui Province, Institutes of Physical Science and Information Technology and School of Internet, Anhui University, Hefei 230601, China, and also with the Fin ChinaAnhui University Joint Laboratory for Financial Big Data Research, Hefei Financial China Information and Technology Company, Ltd., Hefei 230022, China (e-mail: [pengchen@ustc.edu](pengchen@ustc.edu)). Fobao Liu and Jun Zhang are with the National Engineering Research Center for Agro-Ecological Big Data Analysis and Application, Information Materials and Intelligent Sensing Laboratory of Anhui Province, Institutes of Physical Science and Information Technology and School of Internet, Anhui University, Hefei 230601, China (e-mail: [junzhang@ahu.edu.cn](junzhang@ahu.edu.cn)). Bing Wang is with the School of Management Science and Engineering, Anhui University of Finance and Economics, Bengbu 233030, China ([e-mail:](e-mail: wangbing@ustc.edu)[ wangbing@ustc.edu](e-mail: wangbing@ustc.edu)). Digital Object Identiﬁer 10.1109/OJEMB.2024.3367243 multi-scale semantics. This preserves important shallow feature information and allows it to interact with the deep feature, enriching the semantics to some extent. Conclusions: The experimental results demonstrate an accuracy rate of 89.2% in identifying cervical intraepithelial neoplasia while maintaining a lightweight model. This performance exceeds the average accuracy achieved by professional physicians, indicating promising potential for practical application. Utilizing automated colposcopic image analysis and the MFEM-CIN model, this research offers a practical solution to reduce the burden on healthcare providers and imp

FAQs

What problem does the MFEM-CIN framework target?

It targets limited effectiveness of colposcopy-based detection and the lack of resources and specialized physicians for early cervical precancer screening, especially in underserved regions.

How does MFEM-CIN combine CNN and Transformer in the model?

The hybrid design aggregates correlations between local and global features, using multi-scale semantics (via MFEM-CIN components such as MSFE and MSFF) to enrich representations for classification.

What performance does the paper report for identifying cervical intraepithelial neoplasia?

The experimental results report 89.2% accuracy for identifying cervical intraepithelial neoplasia while keeping the model lightweight, exceeding average physician accuracy and indicating practical potential.

Sep 21, 2026

Uploaded by Ava Thompson