Sonoelastomics for Breast Tumor Classification: A Radiomics Approach with Clustering-Based Feature Selection on...

Area of Science:

• Medical Imaging
• Radiology
• Oncology

Background:

• Accurate differentiation of benign and malignant breast tumors is crucial for effective patient management.
• Ultrasound elastography provides information on tissue stiffness, aiding in tumor characterization.

Purpose of the Study:

• To develop and validate a radiomics approach, termed "sonoelastomics," for classifying breast tumors.
• To assess the efficacy of sonoelastomic features in differentiating benign from malignant breast lesions.

Main Methods:

• A high-throughput 364-dimensional feature set was extracted from sonoelastograms.
• Features quantified tumor shape, hardness, and heterogeneity.
• Hierarchical clustering and feature selection metrics were used for dimensionality reduction.

Main Results:

• Seven selected sonoelastomic features achieved high diagnostic performance in a validation set.
• Area under the ROC curve was 0.917, with 88.0% accuracy, 85.7% sensitivity, and 89.3% specificity.
• Sonoelastomics outperformed principal component analysis, deep polynomial networks, and manually selected features.

Conclusions:

• Sonoelastomic features are valuable for differentiating benign and malignant breast tumors.
• This radiomics approach enhances the diagnostic capabilities of ultrasound elastography.
• Further validation in larger cohorts is warranted to establish clinical utility.