Computer-aided diagnosis using morphological features for classifying breast lesions on ultrasound

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Oncology

Background:

• Accurate breast tumor classification is crucial for effective treatment.
• Ultrasound imaging is a widely used modality for breast lesion assessment.
• Computer-aided diagnosis (CAD) systems can enhance diagnostic accuracy.

Purpose of the Study:

• To develop and evaluate a CAD system for breast tumor classification using ultrasound.
• To implement automatic contouring and morphological analysis for feature extraction.
• To assess the system's ability to distinguish between benign and malignant breast tumors.

Main Methods:

• 118 breast lesions (34 malignant, 84 benign) were analyzed.
• Automatic contour extraction from ultrasound images.
• Morphological feature calculation, principal component analysis (PCA), and support vector machine (SVM) classification.
• Performance evaluation using k-fold cross-validation and receiver-operating characteristics (ROC) curve analysis.

Main Results:

• The CAD system achieved an area under the ROC curve of 0.91 with all morphological features.
• Using principal vectors, the area under the ROC curve was 0.90.
• The system demonstrated good classification ability for breast tumors based on morphological information.

Conclusions:

• The developed CAD system effectively differentiates benign from malignant breast tumors.
• The system provides a clinically useful second opinion for breast lesion diagnosis.
• Morphological features are largely setting-independent, ensuring broad applicability across ultrasound machines.