Related Experiment Video
Updated: Jun 6, 2025
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
1.7K
A Machine Learning Model Based on Thyroid US Radiomics to Discriminate Between Benign and Malignant Nodules
Antonino Guerrisi1, Elena Seri2, Vincenzo Dolcetti2
1Radiology and Diagnostic Imaging Unit, Department of Clinical and Dermatological Research, San Gallicano Dermatological Institute IRCCS, Via Elio Chianesi 53, 00144 Rome, Italy.
Cancers
|November 27, 2024
Summary
This study developed a machine learning model using thyroid ultrasound images to accurately classify thyroid nodules as benign or malignant. The model achieved high accuracy, correctly identifying all malignant nodules in external testing.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Thyroid nodules are common, with accurate diagnosis crucial for distinguishing benign from malignant cases.
- Current diagnostic methods like ultrasound and fine needle biopsy have limitations, often operator-dependent.
- Radiomics and machine learning offer advanced tools for improved diagnostic accuracy.
Purpose of the Study:
- To develop and validate a machine learning model for classifying thyroid nodules as benign or malignant using ultrasound images.
- To leverage radiomic features for enhanced diagnostic performance beyond traditional methods.
Main Methods:
- Collected ultrasonography images from 142 subjects (40 malignant, 102 benign) confirmed by histology.
- Applied a radiomic approach with machine learning classifiers (random forests, SVM, k-NN) for binary classification.
- Validated the best performing model on an independent external cohort of 21 patients.
Main Results:
- The best model, an ensemble of random forests, achieved 85% ROC-AUC and 83% accuracy in internal testing.
- In external validation, the model demonstrated 90.5% accuracy, 100% sensitivity, and 86.7% specificity.
- The model successfully identified all malignant nodules and a high proportion of benign nodules in the external cohort.
Conclusions:
- A machine learning model based on radiomic features from ultrasound images can effectively differentiate benign from malignant thyroid nodules.
- The developed random forest ensemble model shows high potential for clinical application in thyroid nodule diagnosis.
- This approach offers a promising, objective tool to aid clinicians in diagnosing thyroid nodules, potentially reducing misdiagnosis.
