Effect of complex wavelet transform filter on thyroid tumor classification in three-dimensional ultrasound

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Ultrasonography is crucial for thyroid nodule differentiation, but visual interpretation faces challenges due to interobserver variability and speckle noise.
• Accurate classification of thyroid nodules is essential for appropriate patient management and treatment strategies.

Purpose of the Study:

• To develop and evaluate an automated system for classifying thyroid tumors in 3D contrast-enhanced ultrasound (CEUS) datasets.
• To enhance the accuracy of malignant versus benign nodule classification by mitigating speckle noise and utilizing advanced feature extraction techniques.

Main Methods:

• A complex wavelet transform-based filter was applied to CEUS images to reduce speckle noise.
• Higher-order spectra features were extracted from processed images.
• A fuzzy classifier was trained and tested using these features for automated nodule classification.

Main Results:

• The automated system achieved a classification accuracy of 91.6% without speckle noise reduction.
• Utilizing the complex wavelet transform filter significantly improved the accuracy to 99.1%.

Conclusions:

• The proposed automated system effectively classifies thyroid tumors in 3D CEUS data.
• The integration of complex wavelet transform filtering and higher-order spectra analysis substantially enhances classification accuracy, offering a promising tool for clinical application.