Related Experiment Video
Updated: Jul 9, 2026
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
1.7K
Self-Attention Mechanisms-Based Laryngoscopy Image Classification Technique for Laryngeal Cancer Detection
Yi-Fan Kang1, Lie Yang2, Yi-Fan Hu3
1Department of Otolaryngology-Head and Neck Surgery, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China.
Head & Neck
|November 11, 2024
Summary
An intelligent laryngeal cancer detection system (ILCDS) using Swin-Transformer shows high accuracy in diagnosing laryngeal cancer (LCA) from laryngoscopic images, outperforming CNN models and human experts.
Area of Science:
- Artificial Intelligence in Medicine
- Medical Imaging Analysis
- Oncology
Background:
- Early diagnosis of laryngeal cancer (LCA) is critical for patient prognosis.
- Developing accurate and sensitive AI models is essential for improving LCA detection rates.
Purpose of the Study:
- To develop and evaluate an intelligent laryngeal cancer detection system (ILCDS) for improved diagnostic accuracy.
- To assess the performance of the ILCDS against traditional AI models and human experts.
Main Methods:
- A dataset of 5768 laryngoscopic images from 1462 patients was utilized.
- The ILCDS was developed using the Swin-Transformer architecture.
- Performance was evaluated on internal and external test sets, compared against CNN models and laryngologists.
Main Results:
- The ILCDS achieved a highest accuracy of 92.78% and an AUC of 0.9732, outperforming six CNN models.
- On external test sets, the ILCDS maintained superior performance with 85.79% accuracy and an AUC of 0.9550.
- The ILCDS demonstrated higher accuracy (92.00%) than professional laryngologists.
Conclusions:
- The ILCDS provides a highly accurate and stable AI solution for laryngeal cancer detection.
- This system has the potential to reduce the diagnostic workload for laryngologists.
