Downgrading Breast Imaging Reporting and Data System categories in ultrasound using strain elastography and...

Area of Science:

• Medical Imaging
• Oncology
• Artificial Intelligence in Medicine

Background:

• Breast ultrasound (US) assessments often lead to unnecessary biopsies for BI-RADS 3 and 4A lesions.
• Accurate classification of breast masses is crucial to avoid invasive procedures.

Purpose of the Study:

• To evaluate if adding elastography strain ratio (SR) and a deep learning-based computer-aided diagnosis (CAD) system can improve the reclassification of BI-RADS 3 and 4A-C categories.
• To determine the potential of these tools in reducing unnecessary breast biopsies.

Main Methods:

• A prospective, multicentre study involving 1049 breast masses (691 benign, 358 malignant) categorized as BI-RADS 3 and 4A-C.
• Patients underwent SR and CAD examinations; histopathology served as the reference standard.
• CAD results were dichotomized as possibly malignant or benign.

Main Results:

• Conventional US led to unnecessary biopsies in 48.6% of benign lesions.
• Reclassification of BI-RADS 4A masses using SR (cut-off <2.90) and CAD reduced unnecessary biopsies by 50.14% (177 of 691).
• The rate of missed malignancies in the reclassified BI-RADS 3 group was 1.72% (9 of 523).

Conclusions:

• Combining SR and CAD shows optimal performance in reclassifying BI-RADS 4A to 3 categories.
• This approach significantly reduces biopsy frequency for BI-RADS 3 and 4A lesions.
• The study demonstrates a substantial reduction in unnecessary biopsies while maintaining an acceptable rate of undetected malignancies.