MRI techniques for detecting focal cortical dysplasia: A systematic review

Abstract

Background: MRI is the cornerstone for detecting and characterising focal cortical dysplasia (FCD), a leading cause of drug-resistant epilepsy. Accurate identification of FCD is critical, as MRI-positive patients have markedly better surgical and clinical outcomes. However, lesion detection can be challenging, particularly in subtle or MRI-negative cases, and a range of MRI techniques has been developed to improve diagnostic performance.

Methods: PubMed, Embase, Scopus, and Web of Science were searched up to April 2025. Diagnostic accuracy studies comparing MRI findings with histopathology or multidisciplinary consensus were retained. 68 studies satisfied eligibility; data extraction was performed, and risk of bias was assessed with QUADAS-2. Marked methodological and outcome heterogeneity precluded meta-analysis, so results were synthesised narratively.

Results: Conventional 1.5T/3T protocols incorporating 3D-T1 and FLAIR were reported to identify most type II lesions, with sensitivities of 50-91 %. At 7T, additional lesions, due partially to the characteristic "black-line" sign, were detected. Quantitative or specialised sequences and post-processing approaches enhanced detection in MRI-negative or type I/III cohorts. Across all patients, machine-learning classifiers yielded sensitivities of 74-93 % but exhibited wide-ranging specificities (34-100 %).

Conclusions: Based on these findings, a tiered diagnostic pathway is recommended: initial evaluation with standard MRI followed, when clinical suspicion persists, by high-field imaging and advanced quantitative or computational methods. Standard MRI detects most type II lesions, but advanced imaging and computational methods improve detection in MRI-negative or subtle cases; real-world implementation requires access, expertise, and standardised validation. Key limitations of the review were study heterogeneity, single-reviewer processes, and lack of consecutively case-sampled studies. The field would benefit from a multi-centre benchmark dataset of operated, histologically confirmed, seizure-free FCD patients, enabling fair head-to-head evaluation of detection methods.