Related Experiment Video
Updated: Jun 20, 2026
A Comprehensive Protocol for Manual Segmentation of the Medial Temporal Lobe Structures
Published on: July 2, 2014
Association between automated brain volumetry and visual classification of hippocampal atrophy in MRI
Fabrício N Garrafiel1, Ricardo B Soder1, Ricardo P Paganin1
1School of Medicine, Pontifical Catholic University of Rio Grande do Sul, Porto Alegre, Rio Grande do Sul, Brazil.
Introduction:
Visual assessment of medial temporal atrophy (MTA) is widely used in Alzheimer's disease (AD) imaging evaluation. However, its subjective nature may contribute to interobserver variability and inconsistencies in hippocampal atrophy classification. Quantitative MRI-based approaches, particularly automated hippocampal volumetry, provide objective measurements that can be statistically adjusted for demographic and anatomical factors. Yet the anatomical and clinical characteristics of individuals showing disagreement between visual MTA assessment and automated volumetric measurements remain poorly characterized. To investigate the relationship between visual MTA ratings and automated hippocampal volumetry, while evaluating interobserver agreement and characterizing discrepant cases between qualitative and quantitative assessments.
Methods:
We analyzed 677 T1-weighted MRI scans from the Alzheimer's Disease Neuroimaging Initiative (ADNI). Hippocampal volumes were obtained using automated segmentation (FreeSurfer), and MTA scores were independently assigned by two experienced neuroradiologists. Interobserver agreement was assessed using Cohen's Kappa coefficient. Associations between hippocampal volumetry and MTA scores were evaluated using correlation and regression analyses adjusted for demographic and intracranial volume variables. Cases with hippocampal volumes beyond ±2 standard deviations from the mean expected for each MTA category were classified as mismatch.
Results:
Interobserver agreement was substantial for both hemispheres (κ = 0.652 left; κ = 0.611 right). Hippocampal volumetry demonstrated a significant inverse correlation with MTA scores (ρ ≈ -0.67). Approximately 10% of individuals (62 of 677) presented discrepant volumetric findings relative to visual classification, predominantly in lower and intermediate MTA categories.
Discussion:
Visual MTA assessment demonstrated relevant variability, particularly in subtle atrophy patterns. Hippocampal volumetry provided complementary quantitative information and may improve objectivity in clinically challenging cases without replacing expert visual interpretation.
