Deep convolutional neural networks in thyroid disease detection: A multi-classification comparison by ultrasonography...

Area of Science:

• Endocrinology
• Medical Imaging
• Artificial Intelligence

Background:

• The thyroid gland, a major endocrine organ, regulates metabolism. Early thyroid disease detection is crucial for reducing mortality.
• Current thyroid disease diagnosis relies heavily on experienced radiologists and pathologists, with potential for human error.
• Deep learning offers a promising approach to mitigate false-positive diagnostic rates in thyroid disease detection.

Purpose of the Study:

• To develop and evaluate a deep learning model for the automatic multi-class diagnosis of thyroid diseases.
• To assess the performance of deep convolutional neural network (CNN) architectures in classifying various thyroid conditions using medical imaging.

Main Methods:

• Utilized two pre-operative medical image modalities: ultrasound and computed tomography (CT) scans.
• Employed a state-of-the-art deep convolutional neural network (CNN) architecture for model development.
• Multi-classified thyroid disease types including normal, thyroiditis, cystic, multi-nodular goiter, adenoma, and cancer.

Main Results:

• The developed deep learning model achieved high accuracy in classifying thyroid diseases from both ultrasound (0.972) and CT scans (0.942).
• Demonstrated unprecedented performance across both medical image datasets.
• The model effectively distinguished among various thyroid disease types.

Conclusions:

• The selected CNN model is adaptable to both ultrasound and CT image modalities for thyroid disease diagnosis.
• The study confirms the feasibility and effectiveness of deep learning models in clinical thyroid disease diagnostics.
• Highlights the potential for widespread application of AI-driven diagnostic tools in endocrinology and radiology.