AI-Rad Companion Brain MR

AI-Rad Companion Brain MR is a medical image analysis software designed to assist clinicians by providing automated segmentation and quantitative volumetric analysis of brain structures and white matter hyperintensities from MR images. Leveraging T1 MPRAGE and T2 FLAIR datasets, it compares findings to normative data and employs non-machine learning based algorithms to track longitudinal changes, such as brain atrophy and lesion progression, to support neurological assessment.

AI-Rad Companion Brain MR VA50 is a software designed for post-processing analysis of MR brain images, aiding clinicians in viewing, evaluating, and quantifying brain structures and white matter hyperintensities (WMH). Its core functionalities, largely inherited from its predicate, include automated segmentation and volumetric analysis of various brain structures using T1 MPRAGE datasets, and of WMH using T1 MPRAGE and T2 weighted FLAIR datasets. The software compares these quantitative measurements against normative data from a healthy, age-matched population, presenting results through numerical reports and visualizations like 3D overlay maps and deviation maps. This updated version, VA50, introduces two new follow-up capabilities: Brain Morphometry follow-up, which calculates atrophy rates for segmented structures between two time points without using machine learning, and White Matter Hyperintensities (WMH) Follow-up. The WMH Follow-up feature identifies and quantifies new or enlarging WMH lesions across two scans and also does not employ machine learning for its core change detection algorithm. Performance validation for the WMH Follow-up feature, conducted on 75 subjects (Multiple Sclerosis and Alzheimer's patients) using 3.0T Siemens scanners, demonstrated robust volumetric agreement of changes (PCC 0.94), voxel-wise spatial overlap of detected changes (Dice 0.50), and WMH change region-wise segmentation accuracy (F1-score 0.69), supporting its utility in tracking longitudinal brain changes.

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 software's performance in segmenting and quantifying white matter hyperintensities vary across different MRI scanner manufacturers and field strengths, beyond the 3.0T Siemens scanners used in the validation study for the follow-up feature?; What was the inter-rater reliability of the ground truth annotations used for validating the WMH follow-up feature, and how might variations in expert opinion affect the perceived accuracy of the algorithm?; Given the normative database is 'age-matched,' how frequently is this database updated, and what is its demographic diversity (e.g., ethnicity, geographic origin) to ensure broad applicability for brain morphometry comparisons?; For the Brain Morphometry follow-up feature, what is the minimum detectable change in atrophy rate that the system can reliably identify, and how does this compare to clinically significant changes for various neurological conditions?; Since the WMH follow-up algorithm does not use machine learning, what specific image processing techniques are employed for identifying new or enlarging lesions, and how robust are they to variations in image quality or registration errors between time points?