Mammography-based radiomic analysis for predicting benign BI-RADS category 4 calcifications

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Breast calcifications require accurate pathological diagnosis.
• Breast Imaging Reporting and Data System (BI-RADS) category 4 calcifications present diagnostic challenges.
• Distinguishing benign from malignant calcifications is crucial for patient management.

Purpose of the Study:

• To develop and validate a radiomic model using mammography for predicting the pathological diagnosis of BI-RADS category 4 calcifications.
• To assess the diagnostic performance of the radiomic model compared to radiologists.
• To evaluate the model's utility in identifying malignant calcifications.

Main Methods:

• Extraction of 8286 radiomic features from mammographic images (CC and MLO views) in 212 patients.
• Development of a radiomic signature using machine learning and integration with clinical risk factors into a nomogram.
• Evaluation of diagnostic performance using the area under the receiver operating characteristic curve (AUC) and DeLong's test.

Main Results:

• The radiomic nomogram, incorporating six radiomic features and menopausal state, achieved an AUC of 0.80 in the validation cohort.
• The radiomic nomogram demonstrated significantly better classification performance than radiologists.
• The model showed strong identification ability for mammography-detected calcifications not visible on ultrasound (MG+/US-).

Conclusions:

• A mammography-based radiomic nomogram is a promising tool for differentiating benign from malignant calcifications.
• This radiomic approach can enhance the accuracy of pathological diagnosis for BI-RADS category 4 calcifications.
• The model offers potential for improved clinical decision-making in breast cancer screening.