Quantra

Quantra™ version 2.2 is a software application designed to aid radiologists in assessing breast tissue composition from digital mammograms by categorizing breast density according to BI-RADS® standards. Its methodology involves analyzing the distribution and texture of parenchymal tissue patterns, an improvement aimed at aligning with the ACR BI-RADS® Atlas (Fifth Edition, 2013) and enhancing estimation accuracy for breast composition.

Quantra™ version 2.2 is a software application designed to assist radiologists in the assessment of breast tissue composition using mammographic images acquired from digital breast x-ray systems. Its primary purpose is to segregate breast density into categories, aiding consistent reporting of BI-RADS® breast composition categories as mandated by certain state regulations, and it reports a result for each subject to assist radiologists. The software analyzes the distribution and texture of parenchymal tissue patterns within these images to estimate breast composition, providing adjunctive information and is not a diagnostic aid. The underlying science of Quantra 2.2 addresses the challenge that fibroglandular, or dense, tissue can obscure or mask abnormal lesions on a mammogram. This version incorporates algorithmic improvements to align with the ACR BI-RADS® Atlas (Fifth Edition, 2013) and enhance breast composition estimation accuracy by focusing on the distribution of glandular tissue rather than solely its density. Non-clinical testing using 230 retrospectively collected 4-view screening negative cases, compared against a five-radiologist consensus for BI-RADS 5th Edition categories, showed high accuracy: for 2D images, 92.9% for fatty (a+b) and 99.1% for dense (c+d) categories; for Tomo (3D center projection) images, 92.0% for fatty and 94.0% for dense categories.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does Quantra's algorithm specifically quantify the "distribution and texture of parenchymal tissue patterns" to differentiate BI-RADS categories, and how has this changed from version 2.1 which calculated volumetric/area density?; What were the specific criteria for the 5 radiologists' consensus in the non-clinical testing, particularly for borderline cases, and what was the inter-observer variability before consensus?; Given the accuracy differences between 2D (99.1% for dense) and Tomo (94.0% for dense) assessments, what specific image characteristics or algorithmic challenges in 3D data contribute to this, and are there plans to address this?; How does Quantra 2.2 ensure robustness and consistent performance across various digital mammography systems and image acquisition protocols, especially concerning the analysis of tissue texture?; The device is "not a diagnostic aid." What further research or features would be needed for Quantra to potentially evolve into a tool that could directly assist in diagnostic decisions related to lesion detection in dense breasts?