Related Experiment Video
Updated: Apr 18, 2026
11:03
High-resolution In Vivo Manual Segmentation Protocol for Human Hippocampal Subfields Using 3T Magnetic Resonance Imaging
Published on: November 10, 2015
10.1K
Training labels for hippocampal segmentation based on the EADC-ADNI harmonized hippocampal protocol
Marina Boccardi1, Martina Bocchetta2, Félix C Morency3
1LENITEM (Laboratory of Epidemiology, Neuroimaging and Telemedicine) IRCCS - Centro S. Giovanni di Dio - Fatebenefratelli, Brescia, Italy.
Alzheimer'S & Dementia : the Journal of the Alzheimer'S Association
|January 25, 2015
Summary
This study provides standardized hippocampal labels from MRI scans for training AI algorithms. These reference labels will enable widespread adoption of accurate Alzheimer's disease and mild cognitive impairment diagnostics.
Area of Science:
- Neuroimaging
- Medical image analysis
- Computational neuroscience
Background:
- The European Alzheimer's Disease Consortium and Alzheimer's Disease Neuroimaging Initiative (ADNI) Harmonized Protocol (HarP) defines manual hippocampal segmentation from MRI.
- Accurate hippocampal segmentation is crucial for training AI tracers and validating automated algorithms.
- Large, representative datasets of segmented hippocampi are needed for reliable AI model development.
Purpose of the Study:
- To generate a comprehensive set of HarP-compliant hippocampal labels.
- To establish a standard of truth for training and certifying AI segmentation algorithms.
- To facilitate the widespread implementation of standardized hippocampal segmentation protocols.
Main Methods:
- Segmentation of 135 volumetric structural ADNI MRI scans (1.5 T and 3 T) by five qualified HarP tracers.
- Ensuring data balance across age, medial temporal atrophy, and scanner manufacturer.
- Centralized quality control and correction to validate HarP compliance of generated labels.
Main Results:
- High interrater reliability (intraclass correlation coefficients of 0.953 for left and 0.975 for right hippocampi).
- Quantified hippocampal volumes and cerebrospinal fluid spaces across control, mild cognitive impairment (MCI), and Alzheimer's disease (AD) groups.
- Significant correlation between hippocampal volumes and atrophy severity (Spearman's ρ < -0.468, P < .0005).
Conclusions:
- The study successfully produced reference hippocampal labels adhering to the HarP standard.
- These publicly released labels will serve as a crucial resource for training and certifying automated segmentation algorithms.
- Widespread adoption of the standard segmentation protocol is anticipated, improving diagnostic accuracy in neurodegenerative diseases.
