Differentiating Mummified Thyroid Nodules From Papillary Thyroid Carcinoma: A Machine Learning Approach Using...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Oncology

Background:

• Distinguishing mummified thyroid nodules (MTNs) from papillary thyroid carcinomas (PTCs) is crucial for appropriate patient management.
• Conventional ultrasound (US) and contrast-enhanced ultrasound (CEUS) provide complementary information for thyroid nodule characterization.
• Radiomics analysis of US and CEUS images offers potential for quantitative feature extraction and improved diagnostic accuracy.

Purpose of the Study:

• To develop and validate a machine-learning model integrating US radiomics, CEUS radiomics, and clinical features for differentiating MTNs from PTCs.
• To assess the diagnostic performance and clinical utility of the developed radiomics model.
• To provide insights into optimal modality combinations for thyroid nodule classification.

Main Methods:

• A retrospective study included 120 PTCs and 84 MTNs.
• Radiomics features were extracted from conventional US and CEUS images.
• Machine learning models, including Logistic Regression and Support Vector Machine, were trained and validated.
• Diagnostic performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA).

Main Results:

• The combined US+CEUS radiomics model achieved an area under the curve (AUC) of 0.936 in the training set and 0.881 in the test set.
• The model, primarily driven by CEUS features, demonstrated superior diagnostic performance compared to models using only US or clinical data.
• Decision curve analysis indicated significant clinical utility of the US+CEUS radiomics model.

Conclusions:

• The integrated US+CEUS radiomics model exhibits high diagnostic value for differentiating PTCs from MTNs.
• CEUS radiomics features play a dominant role in the model's performance.
• Combining clinical features with radiomics did not significantly improve diagnostic accuracy, suggesting careful consideration of modality integration based on disease characteristics.