densitas densityai

densitas densityai™ is an adjunctive software tool that utilizes proprietary image feature-based algorithms to automatically assess breast tissue composition from full field digital mammography and digital breast tomosynthesis images, providing a BI-RADS 5th Edition density category to aid physicians. Its performance has been validated through non-clinical testing against expert radiologist consensus, demonstrating reliability and accuracy in classifying breast density.

densitas densityai™ is a standalone software application designed to assist interpreting physicians in the assessment of breast tissue composition from mammographic images. It operates with compatible full field digital mammography (FFDM) and digital breast tomosynthesis (DBT) systems, analyzing 'for presentation' image data using proprietary, image feature-based algorithms. The software automatically generates a breast density category according to the ACR BI-RADS Atlas 5th Edition, providing this adjunctive information as a DICOM structured report or secondary capture for viewing on mammography workstations, PACS, or transmission to a RIS. This output consists of a DENSITAS breast density grade for each patient, and the system is capable of processing approximately 600 cases per hour. The underlying science involves advanced image processing techniques to identify and quantify image features relevant to breast density, subsequently classifying the breast tissue into one of the four BI-RADS 5th Edition categories. Non-clinical performance validation, conducted without direct human clinical trials, relied on expert radiologist visual assessments of mammography density. This validation included tests for reliability (Pearson’s correlation for bilateral breast measurements), accuracy (for percent density, total area, and breast density grade using the Kappa statistic), reproducibility (Pearson's correlation on repeated analyses), and the expected inverse relationship between breast density and patient age. Compared against a consensus of four expert radiologists on a dataset of 796 cases covering all compatible scanner types, densitas densityai™ achieved a Kappa statistic of 0.87 for the 4-category BI-RADS scale and 0.84 when categories were grouped into fatty (A,B) and dense (C,D) tissue types.

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 the 'image feature-based' assessment methodology of densitas densityai™ differ from the 'area-based' assessment of its predicate, and what is the clinical impact of this change on density classification accuracy across diverse breast morphologies?; Given the Kappa value of 0.87 against a 4-radiologist consensus for the 4-category BI-RADS scale, what was the inter-observer variability among the expert radiologists themselves, and how does this compare to the software's consistency?; The software was validated using 'for presentation' image data. How might its performance be affected by variations in image acquisition protocols or if it were to utilize 'for processing' data from different FFDM and DBT systems?; What specific image features are prioritized by the proprietary algorithms in densitas densityai™ to determine breast density, and how robust are these features against common image artifacts or variations in breast compression?; While the software provides a BI-RADS 5th Edition category, how does it handle borderline cases or images with ambiguous features, and is there a confidence score or uncertainty measure associated with the assigned density category?