Genius AI Detection

Genius AI Detection is a software tool designed to aid radiologists in the interpretation of digital breast tomosynthesis (DBT) examinations by identifying and marking suspicious regions indicative of cancer. It employs deep learning algorithms to analyze DBT images from compatible Hologic systems, providing lesion-specific confidence scores and an overall case score to enhance detection accuracy for masses, architectural distortions, asymmetries, and calcifications.

Genius AI Detection is a software device designed to assist in the interpretation of digital breast tomosynthesis (DBT) images for breast cancer screening. It employs deep learning algorithms to analyze mammographic views from compatible Hologic DBT systems, identifying potential regions of interest such as soft tissue densities (masses, architectural distortions, asymmetries) and calcifications. The system generates detailed CAD results for each detected lesion, including its location, an outline, and a confidence score, along with an overall case score for the entire exam. These findings are packaged into a DICOM Mammography CAD SR object for display on compliant review workstations, allowing physicians to access and evaluate the AI-generated marks concurrently with their own reading process. Performance testing, including a reader study with 17 readers and 390 cases, demonstrated that using Genius AI Detection resulted in an improved average area under the curve (AUC) from 0.794 to 0.825, and an increase in average reader sensitivity for cancer cases from 66.8% to 75.9%. While it slightly increased the average recall rate for non-cancer cases by 2.4% and average case read-time by 5.7 seconds, the device aims to enhance diagnostic performance. Standalone testing confirmed its comparable efficacy across different Hologic image acquisition modes, including standard resolution (1mm slices), high resolution (1mm slices - Clarity HD), and high resolution SmartSlices (6mm - 3DQuorum). The device is specifically compatible with Hologic Selenia Dimensions and Hologic 3Dimensions systems.

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 deep learning model architecture of Genius AI Detection specifically adapt to variations in breast density when identifying lesions?; What types of lesions (e.g., spiculated masses, calcification clusters, architectural distortions) showed the most significant improvement in detection sensitivity with the aid of Genius AI Detection in the reader study?; Considering the 5.7-second increase in average read-time, what is the impact on overall radiologist workflow efficiency in high-volume screening environments, and are there specific interface features that mitigate this?; How does the "Case Score" generated by Genius AI Detection specifically guide radiologists in prioritizing or managing their worklist, and what is its correlation with the likelihood of malignancy?; Are there discernible differences in the AI's performance metrics (e.g., false positive rates, specific lesion detection rates) across the distinct Hologic image reconstruction types it supports (standard resolution 1mm, high resolution 1mm, high resolution 6mm SmartSlices)?