Related Experiment Video
Updated: May 31, 2025
12:09
Network Analysis of the Default Mode Network Using Functional Connectivity MRI in Temporal Lobe Epilepsy
Published on: August 5, 2014
18.0K
Redefining diagnostic lesional status in temporal lobe epilepsy with artificial intelligence.
Ezequiel Gleichgerrcht1, Erik Kaestner2, Reihaneh Hassanzadeh3,4
1Department of Neurology, Emory University, Atlanta, GA 30329, USA.
Brain : a Journal of Neurology
|January 22, 2025
Summary
Artificial intelligence accurately identifies subtle brain atrophy patterns in temporal lobe epilepsy (TLE) patients previously missed by human experts. This AI-driven approach enhances TLE diagnosis, especially for MRI-negative cases, improving patient care.
Area of Science:
- Neuroimaging
- Artificial Intelligence
- Epilepsy Research
Background:
- A significant portion of temporal lobe epilepsy (TLE) patients are classified as 'MRI negative' due to subtle atrophy patterns undetectable by visual MRI assessment.
- This diagnostic uncertainty in MRI-negative TLE patients leads to treatment delays and impacts patient management.
- Quantitative MRI and AI show promise in detecting subtle, TLE-specific atrophy patterns in the temporal and limbic regions.
Purpose of the Study:
- To test the hypothesis that AI, specifically a 3D convolutional neural network (CNN), can accurately detect TLE-specific atrophy patterns in MRI scans.
- To improve the detection of 'lesional' patterns in TLE, particularly in MRI-negative cases.
- To evaluate the AI's performance against traditional methods like support vector machines (SVM).
Main Methods:
- A 3D CNN was applied to 1178 MRI scans from 12 centers to differentiate TLE patients from healthy controls.
- Performance was compared to SVMs using hippocampal and whole-brain volumes.
- Analysis included a subset of surgically treated patients with sustained seizure freedom as a gold standard for TLE confirmation.
Main Results:
- The 3D CNN achieved high accuracy (85.9% ± 2.8%) in differentiating TLE from controls, outperforming SVMs.
- MRI-negative TLE patients were identified with 82.7% ± 0.9% accuracy, validating the AI's ability to detect subtle atrophy.
- Saliency maps highlighted limbic structures (medial temporal, cingulate, orbitofrontal areas) as key for classification, consistent with known TLE atrophy patterns.
Conclusions:
- AI, particularly 3D CNNs, can effectively identify TLE-specific atrophy patterns, even in cases considered MRI-negative by human experts.
- AI-aided diagnosis has the potential to significantly enhance neuroimaging-based TLE diagnosis and redefine the concept of 'lesional' TLE.
- The findings suggest MRI-negative TLE patients fall on a continuum of atrophy common to all TLE cases, detectable by advanced AI.
