Related Experiment Video
Updated: Dec 14, 2025
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
2.2K
Ultrasonographic Thyroid Nodule Classification Using a Deep Convolutional Neural Network with Surgical Pathology.
Soon Woo Kwon1, Ik Joon Choi2, Ju Yong Kang2
1Radiation Medicine Clinical Research Division, Korea Institute of Radiological and Medical Sciences (KIRAMS), Seoul, South Korea.
Journal of Digital Imaging
|July 25, 2020
Summary
This study developed a deep learning model to classify thyroid nodules using ultrasonography, improving diagnostic accuracy for thyroid cancer detection. The AI model achieved high sensitivity and specificity, aiding physicians in diagnosis.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Medicine
- Oncology
Background:
- Thyroid cancer diagnosis commonly relies on ultrasonography and fine-needle aspiration biopsy.
- Current ultrasonography methods for thyroid nodules are limited by subjective interpretation and interobserver variability.
Purpose of the Study:
- To develop and evaluate a convolutional neural network-based classification system for thyroid nodules in ultrasonography.
- To enhance the objectivity and accuracy of thyroid nodule diagnosis using artificial intelligence.
Main Methods:
- A deep learning model was created using transverse and longitudinal ultrasonographic thyroid images from 762 patients.
- Transfer learning with the VGG16 model, data augmentation, and 4-fold cross-validation were employed to train and validate the model.
- The dataset included 325 benign and 437 papillary thyroid carcinoma cases confirmed by surgical biopsy.
Main Results:
- The developed deep learning model demonstrated an average area under the curve of 0.916.
- The model achieved a sensitivity of 0.92, specificity of 0.70, positive predictive value of 0.90, and negative predictive value of 0.75.
- The model effectively classified thyroid nodules, showing promising diagnostic performance.
Conclusions:
- A novel, fine-tuned deep learning model was successfully developed for classifying thyroid nodules in ultrasonography.
- This AI-powered system has the potential to assist physicians in improving the accuracy and consistency of thyroid nodule diagnosis.
- The study highlights the utility of deep learning in overcoming the limitations of subjective interpretation in medical imaging for thyroid cancer detection.
