Related Experiment Video
Updated: May 26, 2026
12:30
A Comprehensive Protocol for Manual Segmentation of the Medial Temporal Lobe Structures
Published on: July 2, 2014
Automatic hippocampal segmentation in temporal lobe epilepsy: impact of developmental abnormalities
Hosung Kim1, Marie Chupin, Olivier Colliot
1Neuroimaging of epilepsy laboratory, McConnell Brain Imaging Center, Montreal Neurological Institute and Hospital, McGill University, Montreal, Quebec, Canada.
Neuroimage
|December 14, 2011
Summary
Automated MRI segmentation for drug-resistant temporal lobe epilepsy (TLE) is challenged by hippocampal malrotation and atrophy. Algorithms showed varying accuracy, with malrotation and atrophy negatively impacting performance, necessitating consideration in future algorithm design.
Area of Science:
- Neurology
- Medical Imaging
- Computational Anatomy
Background:
- Accurate hippocampal segmentation on MRI is crucial for identifying surgical targets in drug-resistant temporal lobe epilepsy (TLE).
- Automated segmentation methods have shown limitations in TLE patients, who often exhibit hippocampal atrophy and developmental malformations (malrotation).
- Understanding the impact of these anatomical variations on automated segmentation performance is essential for improving diagnostic accuracy.
Purpose of the Study:
- To evaluate the performance of three state-of-the-art automated hippocampal segmentation algorithms in TLE patients.
- To assess the influence of hippocampal malrotation and atrophy on the accuracy of these algorithms.
- To compare the accuracy of automated methods against manual segmentation and volumetric measurements.
Main Methods:
- Hippocampal segmentation was performed on 66 TLE patients and 35 healthy controls using SACHA, FreeSurfer, and ANIMAL-multi.
- Malrotation was quantified using 3D models and collateral sulcus analysis; atrophy was assessed via manual volumetry.
- Segmentation accuracy was evaluated using Dice similarity index and surface-based shape mapping, correlated with malrotation and atrophy.
Main Results:
- ANIMAL-multi showed consistent accuracy between patients and controls, while SACHA and FreeSurfer were less accurate in patients.
- Malrotation and atrophy significantly impacted FreeSurfer and ANIMAL-multi accuracy, whereas SACHA was primarily affected by malrotation.
- Automated methods underestimated atrophy magnitude and showed reduced accuracy in lateralizing the seizure focus, particularly with malrotation.
Conclusions:
- Hippocampal malrotation and atrophy negatively affect the performance of current automated segmentation algorithms in TLE.
- Algorithm design must account for these anatomical variants to improve reliability in clinical TLE assessment.
- Further development is needed to enhance the robustness of automated segmentation in the presence of complex hippocampal morphologies.
