Enhancing Diagnostic Efficiency: A Radiomics Approach for Distinguishing Benign and Malignant Breast Lesions Using...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Oncology

Background:

• Breast cancer is a leading cause of cancer mortality in women globally.
• Ultrasound is a common breast cancer detection tool, but its accuracy depends on operator expertise.
• Radiomics offers a potential method to enhance diagnostic accuracy in breast lesion classification.

Purpose of the Study:

• To develop and evaluate a radiomics model for improved differentiation of benign and malignant breast lesions using ultrasound.
• To compare the diagnostic performance of models using BI-RADS features, radiomics features, and a combination of both.
• To assess the interpretability of the developed radiomics model.

Main Methods:

• Retrospective analysis of 316 patients' ultrasound images.
• Extraction of traditional radiomics and BI-RADS (Breast Imaging Reporting & Data System) classification features.
• Development of diagnostic models using Support Vector Machine (SVM), Random Forest (RF), and Logistic Regression (LR) algorithms.
• Performance evaluation using AUC and accuracy, with SHAP for model interpretability.

Main Results:

• The SVM model combining BI-RADS and radiomics features (Group C) achieved the highest performance on the testing set, with an AUC and accuracy of approximately 0.91.
• SHAP analysis identified entropy and variance as the most influential features in the SVM model for Group C.
• The combined feature group demonstrated superior diagnostic efficacy compared to individual feature groups.

Conclusions:

• A Support Vector Machine (SVM) model integrating BI-RADS and radiomics features shows high accuracy in distinguishing malignant from benign breast lesions on ultrasound.
• This radiomics-based approach has the potential to serve as a valuable tool for radiologists, improving diagnostic confidence.
• The findings highlight the utility of combining imaging features and machine learning for enhanced breast cancer diagnosis.