A Hybrid Transfer Learning and Broad Learning Framework for Cervical Cell Classification Using SIPaKMeD Images
International Journal of Medical Toxicology and Forensic Medicine,
Vol. 16 (2026),
1 Dey 2026
,
Page 1-12
https://doi.org/10.22037/ijmtfm.v16.53703
Abstract
Background: Automated cervical cell image classification can support cytological screening by helping identify different cervical cell types. This study proposes a hybrid Transfer Learning and Broad Learning System (TL-BLS) framework for classifying five cervical cell classes.
Methods: We used the SIPaKMeD image collection, comprising 4,049 cervical cell images. Image preprocessing included CLAHE, bilateral filtering, Laplacian sharpening, and unsharp masking. We used Faster R-CNN with an InceptionV2 backbone for transfer-learning-based feature representation, followed by a Broad Learning System for classification. We partitioned the images into training, validation, and testing subsets using a 70%/15%/15% split. This study's main contribution is a combined framework that combines Faster R-CNN with an InceptionV2 backbone for candidate cell-region detection and feature representation, followed by a Broad Learning System (BLS) for five-class cervical cell classification. The proposed architecture integrates transfer-learning-based feature extraction with a broad-learning classifier. We report precision, recall, and F1-score using macro-averaging, the unweighted mean of the corresponding class-wise metrics.
Results: The proposed Hybrid TL-BLS framework correctly classified 578 of 609 test images, achieving 94.91% accuracy and precision, and 94.89% and 94.90% recall and F1-score, respectively.
Conclusion: The proposed framework manifests promising performance for five-class cervical cell-image classification. However, the evaluation is limited to the SIPaKMeD dataset and does not establish patient-level cervical cancer diagnosis or risk prediction. Further validation using independent and clinically representative datasets is required.
- Cervical Cell Classification, SIPaKMeD, Broad Learning System, Transfer Learning, Faster R-CNN, InceptionV2, Medical Image Analysis, Cervical Cytology
How to Cite
References
[1] Stanek-Widera A, Borowczak J, Skiba D, Mickael ME, Łazarczyk M, Maniewski M, et al. Artificial Intelligence in Cervical Cytology: Opportunities and Limitations in Screening, Triage, and Diagnostic Support. Diagnostics. 2026;16(10):1541. [DOI: 10.3390/diagnostics16101541]
[2] Priya S, Panigrahi SK, Bansal V, Arnav A, Raj M, Rai P. Artificial intelligence for cervical cancer screening and diagnosis using Pap smear images: a systematic review. J Obstet Gynecol India. 2026;76(3):192-9. [DOI: 10.1007/s13224-026-02352-4]
[3] Wang X, Wang Q, Ding G, Wang J, Tang Y, Feng Y. Artificial intelligence for colposcopic and cytological image analysis in early cervical cancer detection. iScience. 2026;29(2):114627. [DOI: 10.1016/j.isci.2026.114627]
[4] Valles-Coral MA, et al. Application of artificial intelligence in cervical cytology: a systematic review of deep learning models, datasets, and reported metrics. Front Big Data. 2026;8:1678863. [DOI: 10.3389/fdata.2025.1678863]
[5] Mondal J, Chatterjee R, Kumar Gourisaria M, Sahni M, León-Castro E. Cervical cancer classification using a novel hybrid approach. Front Oncol. 2025;15:1703772. [DOI: 10.3389/fonc.2025.1703772]
[6] Zhang Y, Ning C, Yang W. An automatic cervical cell classification model based on improved DenseNet121. Sci Rep. 2025;15(1):1853. [DOI: 10.1038/s41598-025-87953-1]
[7] Kılıç Eİ, Kılıç Ş. Multi-scale feature integration with enhanced cytomorph for high-accuracy cervical cytology classification. PLoS One. 2026;21(6):e0351063. [DOI: 10.1371/journal.pone.0351063]
[8] Coskun G, Akuner MC, Kaplanoglu E. Robust Cell-Level Classification for Liquid-Based Cervical Cytology Using Deep Transfer Learning: A Multi-Source Study Addressing Scanner-Induced Domain Shifts. Bioengineering. 2026;13(3):289. [DOI: 10.3390/bioengineering13030289]
[9] Fang M, Liao B, Lei X, Wu FX. A systematic review on deep learning based methods for cervical cell image analysis. Neurocomputing. 2024;610:128630. [DOI: 10.1016/j.neucom.2024.128630]
[10] SIPaKMeD: A Segmentation and Image Classification Database for Pap Smear Analysis [Dataset]. University of Ioannina; 2017 [cited 2026 Jan 01]. Available from: https://www.cs.uoi.gr/~marina/sipakmed.html
[11] Ramos-Casallas A, Cardona-Mendoza A, Perdomo-Lara SJ, Rico-Mendoza A, Porras-Ramírez A. Performance evaluation of machine learning models in cervical cancer diagnosis: Systematic review and meta-analysis. Eur J Cancer. 2025;229:115768. [DOI: 10.1016/j.ejca.2025.115768]
[12] Tan SL, Selvachandran G, Ding W, Paramesran R, Kotecha K. Cervical Cancer Classification From Pap Smear Images Using Deep Convolutional Neural Network Models. Interdiscip Sci Comput Life Sci. 2024;16(1):16-38. [DOI: 10.1007/s12539-023-00589-5]
[13] Accuracy-Efficiency Trade-offs in Automated Cervical Cytology Classification: A Comparative Evaluation of ViT, DeiT, PiT, and Swin Transformer on SIPaKMeD. Artif Intell Appl Sci. 2026. [DOI:10.69882/adba.ai.2026021]
[14] Chen CLP, Liu Z. Broad Learning System: An Effective and Efficient Incremental Learning System Without the Need for Deep Architecture. IEEE Trans Neural Netw Learn Syst. 2018;29(1):10-24. [DOI: 10.1109/TNNLS.2017.2716952]
[15] Devi S, Adaickalam V. Modelling a novel approach for cervical cancer prediction using pre-trained network model. Int J Syst Assur Eng Manag. 2025. [DOI: 10.1007/s13198-025-03058-y]
[16] Utomo CP, et al. Optimizing Image Preprocessing for AI-Driven Cervical Cancer Diagnosis. Adv Sustain Sci Eng Technol. 2025;7(1):1-11. [DOI: 10.26877/asset.v7i1.1128]
[17] Hasan MN, Shuvo SB, Ankon MMH, Raju SMTU, Siddique N. TransfusionNet: Framework for cervical cancer detection using deep learning with multi-level fusion. Results Eng. 2025;28:107174. [DOI: 10.1016/j.rineng.2025.107174]
[18] Wang B, Zhao Y, Chen CLP. Hybrid Transfer Learning and Broad Learning System for Wearing Mask Detection in the COVID-19 Era. IEEE Trans Instrum Meas. 2021;70:5004210. [DOI: 10.1109/TIM.2021.3069844]
- Abstract Viewed: 60 times
- 53703 Downloaded: 26 times