CorInsights MRI

CorInsights MRI is an automated software that quantifies brain structure volumes from T1-weighted MR images using an atlas-based segmentation methodology and comparison to an internal normative reference database. It aims to provide clinicians and researchers with objective, quantitative morphometric data to assist in evaluating structural brain changes in adults aged 45 to 95, with its efficacy supported by validation studies demonstrating high accuracy and reproducibility across diverse clinical cohorts and imaging parameters.

CorInsights MRI is a fully automated medical image processing software designed to perform automatic labeling, visualization, and volumetric quantification of identifiable brain structures using DICOM formatted T1-weighted magnetic resonance images. The software produces a PDF report containing morphometric measurements, which are compared to age and gender-matched reference percentile data, and an annotated DICOM image displaying anatomical structure boundaries with color overlays. Its internal architecture features a proprietary process involving artifact correction, atlas-based segmentation, volume calculation, and report generation, complemented by automated quality control measures such as image header checks and tissue contrast verification. CorInsights MRI is intended for adults aged 45 to 95 and serves as a support tool for medical professionals and clinical researchers in the assessment of structural MRIs. The device's performance was substantiated through comprehensive testing. Accuracy evaluations, using manually segmented ground truth data, demonstrated high performance for hippocampal volume (IntraClass Correlation coefficient [ICC] 0.95, DICE coefficient 83%), total gray matter volume (ICC 0.99, DICE 95%), intracranial volume (ICC 0.89, DICE 95%), and ventricular volumes (ICC 0.98, DICE ~87-88%). Reproducibility was also high, with an average test-retest ICC of 0.97. A normative reference database was established from 600 cognitively normal individuals (ages 42-95), carefully screened for amyloid pathology and other neurological confounders. Further validation involved over 1,400 scans from more than 1,100 individuals, encompassing a wide range of clinical diagnoses (cognitively normal, Mild Cognitive Impairment, Alzheimer's disease variants, non-Alzheimer's dementias) and various MRI scanner models (Siemens, GE, Philips) and field strengths (1.5T, 3T), confirming that the software's volumetric measurements align with established findings in peer-reviewed literature for these patient populations.

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 atlas-based segmentation algorithm adapt to significant anatomical variability observed in advanced stages of neurodegenerative diseases or across diverse populations not extensively represented in the normative database?; What are the specific artifact correction methods implemented in CorInsights MRI, and what is their quantified impact on segmentation accuracy and volumetric measurements under various common MR artifact conditions (e.g., motion, Gibbs ringing, intensity inhomogeneity)?; Considering the manual ground truth segmentations were generated by 'neuroanatomy experts,' what was the protocol for ensuring inter-expert consistency, and what were the measured inter-rater reliability scores for these ground truth datasets?; What is the longitudinal stability of CorInsights MRI's volumetric measurements, and how does it account for potential scanner upgrades or protocol changes when tracking individual patient changes over time?; How does CorInsights MRI's performance, particularly in measuring subtle volumetric changes in structures like the hippocampus, compare when using MR images acquired with accelerated sequences versus non-accelerated sequences, beyond general validation across both types?