A first evaluation of breast radiological density assessment by QUANTRA software as compared to visual classification

Area of Science:

• Radiology
• Medical Imaging
• Oncology

Background:

• Breast density is a key factor in breast cancer risk and mammogram effectiveness.
• Personalized screening strategies can be informed by breast density assessments.
• Variability in visual density assessment poses challenges for consistent clinical application.

Purpose of the Study:

• To evaluate the reproducibility of visual breast density assessment by radiologists.
• To compare computer-assisted breast density measurements with visual assessments.
• To establish a reliable method for computer-assessed breast density in clinical practice.

Main Methods:

• Eleven experienced radiologists visually classified 418 digital mammograms using BI-RADS categories.
• The QUANTRA software was used for computer-assisted breast density assessment.
• Statistical analysis, including kappa values and cut-off analysis, was performed to compare methods.

Main Results:

• Visual density assessment showed satisfactory reproducibility (kappa 0.620-0.851) but significant inter-reader variability (25.1-50.5% dense breasts).
• QUANTRA software yielded systematically lower density percentages than visual assessment.
• A QUANTRA cut-off of ≤22.0% accurately predicted visual classification into dense and non-dense categories (89.0% overall accuracy).

Conclusions:

• Computer-assessed breast density is highly reproducible and preferable to subjective visual assessment.
• The QUANTRA software, with a threshold of ≤22.0%, can reliably differentiate dense from non-dense breasts.
• Further comparative studies are needed to refine computer-assessed density for clinical use.