AI Tool Improves Detection of Focal Cortical Dysplasia in Epilepsy Patients

A groundbreaking study published in JAMA Neurology has shown that a newly developed artificial intelligence (AI) tool significantly improves the detection of focal cortical dysplasia (FCD) — a leading cause of drug-resistant epilepsy often missed on standard MRI scans.

The research, conducted under the Multicenter Epilepsy Lesion Detection (MELD) consortium, involved MRI data from 23 epilepsy centers worldwide. The study included 703 individuals with epilepsy due to FCD and 482 control participants.

Using a context-aware graph neural network (MELD Graph), the AI tool analyzes surface-based MRI features while incorporating spatial relationships across brain regions. This approach addresses long-standing challenges in detecting subtle FCD lesions, especially in patients deemed “MRI-negative.”

Key Findings:

• Achieved 81.6% sensitivity for confirmed FCD in patients who became seizure-free one year after surgery.
• Detected 63.7% of lesions in MRI-negative patients that radiologists had previously missed.
• Demonstrated a positive predictive value (PPV) of ~67% on internal testing and 76% PPV in an external validation cohort — nearly doubling the performance of baseline algorithms.
• Generated interpretable clinical reports, including lesion location, size, features, and confidence scores, supporting review by epilepsy teams.

Lead investigators highlight that this open-access AI tool could serve as a valuable clinical aid, helping neuroradiologists and epilepsy specialists detect subtle abnormalities, improve presurgical planning, and increase the likelihood of successful surgical outcomes for patients with drug-resistant epilepsy.

The study emphasizes that while the tool is highly promising, further prospective clinical trials are needed to confirm its utility in everyday practice.

For more information, please refer to the full article in JAMA Neurology: Link to study.