Longitudinal quantitative MRI feature analysis in paediatric focal cortical dysplasia: a retrospective longitudinal cohort study.
Authors
Affiliations (5)
Affiliations (5)
- Developmental Neurosciences, UCL Great Ormond Street Institute of Child Health, London, UK. [email protected].
- Developmental Neurosciences, UCL Great Ormond Street Institute of Child Health, London, UK.
- Great Ormond Street Institute of Child Health, University College London, London, UK.
- School of Biomedical Engineering - Imaging Sciences, Kings College London, London, UK.
- Developmental Neurosciences, UCL Great Ormond Street Institute of Child Health, London, UK. [email protected].
Abstract
Focal cortical dysplasia (FCD) is a common cause of treatment-resistant epilepsy and can be difficult to detect on MRI. Although traditionally considered static, lesion visibility may vary across scans. We assessed whether MELD-derived lesional imaging features differ across serial MRI in histologically confirmed paediatric FCD type II. Twenty-one patients with histopathologically confirmed FCD type II underwent paired T1-weighted MRI; 13 also had paired FLAIR imaging. Images were processed using FastSurfer and a modified MELD feature-extraction pipeline. Lesion-mask feature values were compared with anatomically matched contralateral regions, and time-corrected longitudinal lesion-contralateral differences were calculated. Blinded qualitative neuroradiological review was available for 20/21 subjects. FCD lesions demonstrated quantitative feature differences from contralateral cortex. In the FLAIR cohort, longitudinal lesion-contralateral differences were consistently positive across sampled grey and white matter depths and remained significant after false discovery rate correction (mean paired differences 0.607-0.737; all pFDR = 0.009). Protocol-restricted FLAIR sensitivity analyses preserved the direction of effect, although restricted subgroups were underpowered. In T1-weighted MRI, longitudinal lesion-contralateral differences were positive across grey matter intensity features, greatest at 75% cortical thickness (mean paired difference 0.185, 95% CI 0.060-0.309, p = 0.006), but no T1 feature survived false discovery rate correction. Neuroradiological review identified possible visually appreciable interval change in a minority of cases, with most apparent change attributed to acquisition-related variation or artefact. Quantitative MRI-derived FCD features can vary across serial paediatric MRI. These findings demonstrate longitudinal imaging feature variability, most consistently in FLAIR-derived measures, with implications for repeat imaging and machine learning derived lesion detection.