Overview of Updates, Modeling, Validation, and Clinical Impact
The NeuroQuant normative database is designed to compare individual brain measurements with those from a large cohort of healthy people matched by age and sex. Volumetric AI tools, like NeuroQuant®, translate raw MRI volumes into percentiles—similar to children’s growth charts—enabling clinicians to interpret data quickly and accurately. This summary highlights how the updated database was built, its advanced statistical modeling, validation processes, and the implications for clinical practice and research.
Building a Comprehensive Reference Population
Cortechs.ai compiled over 16,400 MRI scans from cognitively normal individuals aged 3 to 100 years, with near-equal gender distribution. Data sources included public studies, partner collaborations, and proprietary datasets from the U.S. and abroad. All scans came from individuals free of neurological or psychiatric disorders, confirmed by clinical history or study criteria. MRI data from 1.5 T and 3 T scanners were processed using NeuroQuant’s automated segmentation, normalizing regional volumes by intracranial volume (ICV). Of the scans, 5,200 were used to fit percentile curves, and 11,200 were reserved for validation, ensuring robust model development without overfitting.
Advanced Statistical Modeling: The LMS Approach
Previous NeuroQuant versions used generalized additive models (GAM), which assumed symmetrical, constant variance. The updated database employs the Lambda–Mu–Sigma (LMS) method via generalized additive models for location, scale, and shape (GAMLSS). LMS models:
- Lambda (L): Age-dependent skewness, normalizing the distribution
- Mu (M): Age-dependent median, representing typical volume
- Sigma (S): Age-dependent variability around the median
This method fits smooth L, M, and S curves across age, capturing non-linear growth and atrophy patterns—such as hippocampal volumes peaking in young adulthood and declining with age. Sex differences are accounted for by adding sex as a covariate. Outliers beyond four standard deviations are removed to ensure accuracy. Percentiles (1st–99th) are computed with NeuroQuant and reports highlight anything below the 5thor above the 95th percentiles. LMS transforms volumes into z‑scores, providing precise, interpretable metrics.
Validation and Quality Control
The updated model underwent extensive validation:
- Visual inspection: Percentile curves accurately matched data spread across ages.
- Z‑score distribution: LMS-derived z‑scores were normally distributed, confirming calibration.
- External overlay: Less than 2% of volumes fell outside the 1st/99th percentiles, with no age-dependent bias.
- Clinical validation: Percentile shifts tracked disease progression (e.g., Alzheimer’s), distinguishing normal aging from pathology.
- Regulatory compliance: Validation maintained FDA clearance and CE marking.
Routine use remains efficient, as percentile lookups are computationally trivial, and workflow for clinicians is unchanged.
Clinical Impact
With over 350,000 scans processed in 2025, updating the normative database ensures NeuroQuant remains sensitive to subtle changes in disease progression and reflects evolving population demographics.
Key clinical and research impacts include:
- Improved diagnostic confidence: Accurate percentiles enable early detection and monitoring of neurological conditions.
- Broad applicability: Norms for ages 3–100 support pediatric, adult, and geriatric assessments.
- Standardized research framework: Enables effect size quantification and collaborative studies.
- Personalized medicine: Age‑ and sex‑matched percentiles allow tailored treatment decisions.
Conclusion
The updated NeuroQuant normative database marks a significant advance in quantitative neuroimaging. Built from a diverse, carefully screened population and powered by robust LMS modeling, it provides accurate, unbiased, and clinically meaningful percentiles. Seamless integration ensures efficient workflow, and ongoing expansions promise even more comprehensive reference data, empowering clinicians and researchers to deliver improved patient care.
