Predictive quantitative sonographic features on classification of hot and cold thyroid nodules

Area of Science:

• Medical Imaging
• Radiology
• Artificial Intelligence in Medicine

Background:

• Thyroid nodules are common, and distinguishing between hot (hypofunctioning) and cold (hyperfunctioning) nodules is crucial for diagnosis.
• Conventional ultrasound imaging can be subjective in characterizing thyroid nodules.

Purpose of the Study:

• To investigate the potential of ultrasound imaging combined with computer-aided diagnosis (CAD) for classifying hot and cold thyroid nodules.
• To assess the efficacy of textural and morphological analysis in differentiating nodule function.

Main Methods:

• Evaluation of 42 hot and 42 cold thyroid nodules using a CAD system.
• Extraction of 49 sonographic features (9 morphological, 40 textural).
• Classification using a support vector machine (SVM) classifier.

Main Results:

• The combination of morphological and textural features achieved an AUC of 0.992 in the training set.
• The proposed CAD model successfully classified hot and cold thyroid nodules with an AUC of 0.948 in the test set.

Conclusions:

• A CAD method employing textural and morphological features can reliably distinguish between hot and cold thyroid nodules using 2D sonography.
• This CAD approach can serve as a valuable supplementary tool in clinical practice to enhance radiologists' nodule characterization capabilities.