NeuroQuant

NeuroQuant is an automated medical image processing software designed for the volumetric quantification of segmentable brain structures and lesions from DICOM-formatted 3D T1-weighted and T2 FLAIR MR images. It utilizes a proprietary Dynamic Atlas™ technology for segmentation, generating morphometric reports that compare measured volumes to reference percentile data to support clinicians in assessing structural MRIs.

NeuroQuant is an automated medical image processing software designed for labeling, visualization, and volumetric quantification of segmentable brain structures and lesions from MR images. Its primary purpose is to automate the manual process of identifying and measuring these structures, providing quantitative data that can be compared to normative reference percentiles based on age and gender for patients from 3 to 100 years old. The LesionQuant module, an extension of NeuroQuant, specifically utilizes T2 FLAIR MR images in conjunction with 3D T1 images to improve the identification and quantification of brain lesions, which are often associated with T2 FLAIR hyperintensities and typically defined as volumes greater than 4 mm³. The outputs, including various morphometric reports (such as General Morphometry, Age Related Atrophy, and Lesion Reports) and color-coded segmentations, are intended to assist neurologists, radiologists, and neuroradiologists as a support tool in their clinical assessment when used with other clinical findings, but do not provide a diagnosis on their own. The system operates by processing DICOM-formatted 3D T1-weighted MRI series (and compatible 2D or 3D T2 FLAIR series for the LesionQuant module) through a proprietary automated internal pipeline. This pipeline performs image filtering, artifact correction, segmentation, volume calculation, and report generation. A core component of NeuroQuant's methodology is its proprietary Dynamic Atlas™ technology, an atlas-based segmentation method. Performance evaluations against expert manual segmentations showed Dice coefficients in the range of 80-90% for major subcortical brain structures and 75-85% for major cortical regions. For lesion segmentations using 3D T1 and T2 FLAIR MRI scan pairs, the Dice coefficient exceeded 80%. Reproducibility studies on repeated 3D T1 MRI scans indicated mean percentage absolute volume differences for major subcortical structures in the range of 1-5%. For brain lesion segmentation reproducibility, the mean absolute lesion volume difference was less than 0.25cc, and the mean percentage lesion absolute volume difference was less than 2.5%, with a Pearson correlation coefficient of 0.99.

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 'Dynamic Atlas TM' technology adapt to the wide age range (3-100 years) and various anatomical presentations, especially in the presence of significant pathologies like tumors or surgical resections, which are listed as factors affecting results?; What specific image filtering and artifact correction algorithms are employed prior to segmentation, and how are their efficacies validated across different MR scanner vendors and protocols?; The LesionQuant module defines a lesion based on brightness relative to gray/white matter on T2 FLAIR and a size threshold (>4 mm3, configurable). How does the system handle lesions with atypical signal intensities or those near the configurable size threshold, and what is the impact of threshold adjustment on sensitivity and specificity for lesion detection?; Given that performance (Dice coefficient) for cortical regions (75-85%) is generally lower than for subcortical structures (80-90%), what are the primary challenges in segmenting cortical regions, and are there ongoing efforts or specific algorithmic components designed to improve accuracy in these areas?; The user manual mentions 'alternative reference data (NOT FOR CLINICAL USE)' including MCI and AD databases. What are the specific characteristics, cohort sizes, and limitations of these alternative databases, and what types of research questions are they intended to help address, given their non-clinical use restriction?