Search

Objective This retrospective, cross‐sectional study evaluated the feasibility and potential benefits of incorporating deep‐learning on structural magnetic resonance imaging (MRI) into planning stereoelectroencephalography (sEEG) implantation in pediatric patients with diagnostically complex drug‐resistant epilepsy. This study aimed to assess the degree of colocalization between automated lesion detection and the seizure onset zone (SOZ) as assessed by sEEG. Methods A neural network classifier was applied to cortical features from MRI data from three cohorts. (1) The network was trained and cross‐validated using 34 patients with visible focal cortical dysplasias (FCDs). (2) Specificity was assessed in 20 pediatric healthy controls. (3) Feasibility of incorporation into sEEG implantation plans was evaluated in 34 sEEG patients. Coordinates of sEEG contacts were coregistered with classifier‐predicted lesions. sEEG contacts in seizure onset and irritative tissue were identified by clinical neurophysiologists. A distance of

OBJECTIVE Focal cortical dysplasia (FCD) lesion detection and subtyping remain challenging on conventional MRI. New diffusion models such as the spherical mean technique (SMT) and neurite orientation dispersion and density imaging (NODDI) provide measurements that potentially produce more specific maps of abnormal tissue microstructure. This study aims to assess the SMT and NODDI maps for computational and radiological lesion characterization compared to standard fractional anisotropy (FA) and mean diffusivity (MD). METHODS SMT, NODDI, FA, and MD maps were calculated for 33 pediatric patients with suspected FCD (18 histologically confirmed). Two neuroradiologists scored lesion visibility on clinical images and diffusion maps. Signal profile changes within lesions and homologous regions were quantified using a surface‐based approach. Diffusion parameter changes at multiple cortical depths were statistically compared between FCD type IIa and type IIb. RESULTS Compared to fluid‐attenuated inversion recovery (FLAIR) or T1‐weighted imaging, lesions conspicuity on NODDI intracellular volume fraction (ICVF) maps was better/equal/worse in 5/14/14 patients, respectively, while on SMT intra‐neurite volume fraction (INVF) in 3/3/27. Compared to FA or MD, lesion conspicuity on the ICVF was better/equal/worse in 27/4/2, while on the INVF in 20/7/6. Quantitative signal profiling demonstrated significant ICVF and INVF reductions in the lesions, whereas SMT microscopic mean, radial, and axial diffusivities were significantly increased. FCD type IIb exhibited greater changes than FCD type IIa. No changes were detected on FA or MD profiles. SIGNIFICANCE FCD lesion‐specific signal changes were found in ICVF and INVF but not in FA and MD maps. ICVF and INVF showed greater contrast than FLAIR in some cases and had consistent signal changes specific to FCD, suggesting that they could improve current presurgical pediatric epilepsy imaging protocols and can provide features useful for automated lesion detection.

The high gray-white matter contrast and spatial resolution provided by T1-weighted magnetic resonance imaging (MRI) has made it a widely used imaging protocol for computational anatomy studies of the brain. While the image intensity in T1-weighted images is predominantly driven by T1, other MRI parameters affect the image contrast, and hence brain morphological measures derived from the data. Because MRI parameters are correlates of different histological properties of brain tissue, this mixed contribution hampers the neurobiological interpretation of morphometry findings, an issue which remains largely ignored in the community. We acquired quantitative maps of the MRI parameters that determine signal intensities in T1-weighted images (R1 (=1/T1), R2 *, and PD) in a large cohort of healthy subjects (n = 120, aged 18-87 years). Synthetic T1-weighted images were calculated from these quantitative maps and used to extract morphometry features-gray matter volume and cortical thickness. We observed significant variations in morphometry measures obtained from synthetic images derived from different subsets of MRI parameters. We also detected a modulation of these variations by age. Our findings highlight the impact of microstructural properties of brain tissue-myelination, iron, and water content-on automated measures of brain morphology and show that microstructural tissue changes might lead to the detection of spurious morphological changes in computational anatomy studies. They motivate a review of previous morphological results obtained from standard anatomical MRI images and highlight the value of quantitative MRI data for the inference of microscopic tissue changes in the healthy and diseased brain. Hum Brain Mapp 37:1801-1815, 2016. © 2016 The Authors. Human Brain Mapping Published by Wiley Periodicals, Inc.