Related Experiment Video
Updated: Aug 5, 2026
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
A Thyroid Nodule Differentiation Model for Benign-Malignant Identification by Fusing Transfer Learning and Gradient
Bo Li1, Tingxue Li2, Hao Ju3
1School of Automation and Electrical Engineering, Shenyang Ligong University, Shenyang, 110159, China.
Journal of Imaging Informatics in Medicine
|July 29, 2026
Summary
This study introduces an optimized AI model for diagnosing thyroid nodules using ultrasound images. The AI achieves 95% accuracy, improving upon traditional methods for better clinical decisions.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Accurate thyroid nodule diagnosis is crucial for treatment planning.
- Traditional ultrasound diagnosis suffers from subjectivity.
- Existing AI models struggle with generalization in small-sample medical data.
Purpose of the Study:
- To develop an optimized computer-aided diagnosis (CADx) model for thyroid nodules.
- To address subjectivity and generalization issues in thyroid nodule diagnosis.
- To improve the accuracy and efficiency of thyroid nodule identification.
Main Methods:
- Utilized ResNet50 with transfer learning for high-level semantic feature extraction from ultrasound images.
- Implemented a two-stage diagnosis architecture.
- Integrated an optimized XGBoost classifier for efficient classification.
Main Results:
- The optimized CADx model achieved a diagnostic accuracy of 95% on a multi-center clinical ultrasound dataset.
- Demonstrated superior performance in precision, recall, and F1-score compared to the original hybrid model and single algorithms.
- Showcased promising diagnostic performance on an internal retrospective dataset.
Conclusions:
- The proposed AI framework offers an objective and reliable diagnostic reference for clinicians.
- It holds significant clinical value in reducing unnecessary fine-needle aspiration biopsies.
- The model enhances the efficiency of early thyroid disease diagnosis, though external validation is needed.
