Automated Breast Density Measurements From Chest Computed Tomography Scans

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Quantitative Imaging Analysis

Background:

• Accurate breast density assessment is crucial for mammography interpretation and breast cancer risk stratification.
• Current methods for breast density quantification often rely on subjective visual assessment or manual segmentation, which can be time-consuming and prone to inter-observer variability.
• Chest computed tomography (CT) scans, routinely acquired for other clinical indications, contain information about breast tissue composition.

Purpose of the Study:

• To develop and validate an automated algorithm for quantifying percent breast density from clinical chest CT scans.
• To assess the accuracy of the automated method against a reference standard of manual segmentation.
• To evaluate the agreement between automated CT-based density estimates and subjective radiologist assessments.

Main Methods:

• A naïve Bayesian classifier was developed using gray-level intensities and spatial relationships from CT scans of 10 Hodgkin lymphoma patients.
• The algorithm was trained and validated on CT scans from a total of 85 Hodgkin lymphoma patients, using consensus manual segmentation of fibroglandular tissue as the reference.
• Pixel-level accuracy was assessed using true and false positive fractions, while patient-level agreement was evaluated using the concordance correlation coefficient (ρc) and Kendall's τb.

Main Results:

• The automated method achieved a pixel-level true positive fraction of 82.7% for identifying fibroglandular tissue, with a false positive fraction of 9.2%.
• Patient-level agreement between the automated density estimate and the reference dataset was high (ρc = 0.93).
• Agreement between the automated CT density estimates and subjective ACR BI-RADS assessments by a radiologist was also substantial (τb = 0.77).

Conclusions:

• Automated quantification of percent breast density from clinical chest CT scans is feasible.
• The developed algorithm demonstrates high accuracy and reliable agreement with established methods for breast density assessment.
• This automated approach offers a potential tool for leveraging existing CT data for breast density evaluation.