Related Experiment Video
Updated: Jun 5, 2025
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
1.7K
CerviFusionNet: A multi-modal, hybrid CNN-transformer-GRU model for enhanced cervical lesion multi-classification
Yuyang Sha1, Qingyue Zhang2,3, Xiaobing Zhai1
1Center for Artificial Intelligence Driven Drug Discovery, Faculty of Applied Sciences, Macao Polytechnic University, Macau SAR 999078, China.
Iscience
|December 5, 2024
Summary
This study introduces CerviFusionNet, an AI tool for cervical lesion screening using colposcopy images. It improves accuracy by analyzing dynamic visual changes, aiding early detection in women's health.
Area of Science:
- Gynecology
- Medical Imaging
- Artificial Intelligence
Background:
- Cervical lesions present a global health risk, with colposcopy effectiveness varying by clinician expertise.
- Artificial intelligence (AI) shows promise for cervical lesion screening using colposcopy images, but faces challenges like algorithm performance and data limitations.
Purpose of the Study:
- To establish a comprehensive multi-modal colposcopy dataset for AI development.
- To develop and evaluate an AI model, CerviFusionNet, for improved cervical lesion detection.
Main Methods:
- Created a multi-modal dataset from 2,273 HPV+ patients, including images and diagnostic data.
- Developed CerviFusionNet, a hybrid CNN-Transformer architecture with a temporal module for dynamic analysis.
- Evaluated CerviFusionNet against existing methods using the established dataset.
Main Results:
- CerviFusionNet achieved high accuracy and efficiency in cervical lesion screening.
- The temporal module effectively captured dynamic acetic acid reaction changes, enhancing performance.
- The developed dataset supports advanced AI research in cervical cancer screening.
Conclusions:
- CerviFusionNet offers a promising AI solution for objective and efficient cervical lesion screening.
- The multi-modal dataset and novel architecture address key limitations in current AI approaches.
- This work contributes to advancing women's health through AI-powered colposcopy analysis.
