Related Experiment Video
Updated: Mar 7, 2026
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
2.4K
Deep learning-based multi-class classification of thyroid disorders on Tc-99m scintigraphy using modified
Hafiz Muhammad Usman Ghani1, Javed Khan2, Naimat Ullah Khan3
1Department of Physics, University of Science & Technology Bannu, Bannu, Pakistan.
Digital Health
|March 6, 2026
Summary
This study introduces an AI system using DenseNet-201 for accurate thyroid disorder diagnosis, achieving 91.48% accuracy. The automated tool assists physicians in identifying seven thyroid conditions, improving diagnostic efficiency.
Area of Science:
- Medical Imaging and Diagnostics
- Artificial Intelligence in Healthcare
- Endocrinology
Background:
- Thyroid gland dysfunction poses significant health risks, necessitating accurate and timely diagnosis.
- Early detection of thyroid disorders is crucial for effective patient recovery and management.
- Current diagnostic methods may benefit from enhanced accuracy and efficiency through technological integration.
Purpose of the Study:
- To develop an automated system for assisting physicians in the clinical diagnosis of thyroid gland disorders.
- To leverage deep learning, specifically the DenseNet-201 model, for classifying various thyroid conditions.
- To enhance the accuracy and efficiency of thyroid disorder diagnosis through an AI-powered approach.
Main Methods:
- Utilized the DenseNet-201 deep neural network model with transfer learning capabilities.
- Modified the fully connected and classification layers of DenseNet-201 for specific thyroid condition classification.
- Trained and evaluated the model on seven categories: cold nodule, hot nodule, multi-nodular goiter, nodular goiter, thyroiditis, toxic diffuse goiter, and normal thyroid conditions.
Main Results:
- Achieved high performance metrics, including 91.48% accuracy, 98.58% specificity, 91.57% precision, 91.48% sensitivity, and a 0.988 Area Under the Curve (AUC).
- The kappa coefficient, measuring agreement with expert diagnosis, was 0.9148, indicating strong concordance.
- Demonstrated superior performance compared to contemporary methods in key diagnostic metrics.
Conclusions:
- The developed automated system shows significant potential for clinical application in thyroid disorder diagnosis.
- The AI model effectively assists physicians by providing accurate classification of various thyroid conditions.
- The high accuracy and strong agreement with expert diagnoses suggest the system can improve patient care and outcomes.
