A comprehensive predictive model study: ultrasound viscosity imaging in thyroid nodules.

Area of Science:

• Radiology
• Medical Imaging
• Oncology

Background:

• Thyroid nodules are common, requiring accurate differentiation between benign and malignant types.
• Conventional ultrasound (US) features have limitations in definitively diagnosing thyroid nodules.
• Ultrasound viscosity imaging offers novel metrics for assessing tissue properties.

Purpose of the Study:

• To develop a predictive model integrating clinical factors, conventional US features, and US viscosity imaging metrics.
• To enhance diagnostic accuracy for thyroid nodules.
• To explore the role of viscosity metrics in differentiating benign from malignant thyroid nodules.

Main Methods:

• A cohort of 215 patients with thyroid nodules was studied, split into training (70%) and testing (30%) sets.
• Three models were developed: ModA (clinical factors), ModB (clinical + conventional US), and ModC (clinical + conventional US + US viscosity imaging).
• Logistic regression and receiver operating characteristic (ROC) curve analysis were used to evaluate diagnostic efficacy.

Main Results:

• Echogenicity, viscosity minimum (Vmin), and viscosity mean (Vmean) were significant predictors of malignancy.
• Model C (including viscosity imaging) achieved the highest Area Under the Curve (AUC) of 0.945 in the testing set.
• ModC demonstrated superior predictive accuracy compared to models with only clinical factors (AUC=0.777) or clinical factors plus conventional US (AUC=0.922).

Conclusions:

• Ultrasound viscosity imaging, when combined with clinical and conventional US data, significantly improves the predictive performance for thyroid nodule differentiation.
• This integrated approach provides a reliable tool to support early and accurate diagnosis of thyroid nodules.
• Viscosity metrics play a crucial role in distinguishing benign from malignant thyroid lesions.