Abstracts

Rationale: Identification of variant patterns of language dominance requires large populations because of patient heterogeneity. We established a consortium of imaging centers to collect functional imaging data using common paradigms and similar acquisition parameters. We aimed to establish similarity of findings across sites, and establish data driven methods to identify sub-patterns of language processing from pooled data. We also aimed to develop data driven means for data segmentation independent of a priori notions and bias inherent in region of interest and visual analysis. Methods: 122 children, (64 normal control; 58 with localization related epilepsy (LRE). Mean age 11.1 yrs; range 4-19 yrs) across five sites using EPI BOLD fMRI and a word definition decision task adjusted for age (reverse speech control) at 1.5 and 3T. Data were placed in an image data repository for central standardized processing; clinical measures were entered in a study Web site repository. Data were processed using FSL. After normalization to the MNI atlas data were segregated using an eigenvector parsing followed by cluster and partial component analysis. Group activation maps form subgroups were generated (p<0.05). Data were examined for evidence subject differences across groups, scanner effect (both Fisher Exact Test) or site effect (Good Fit Test). Results: There were no scanner (p=0.7) or site effects (p=0.58). Three patterns of activation were identified: 1) the typical distributed network expected for task in left inferior frontal gyrus (Broca’s) and along left superior temporal gyrus (Wernicke’s) (60 controls, 35 patients); 2) a variant left dominant pattern with greater activation in IFG, mesial left frontal lobe, right cerebellum (3 controls, 15 patients); and 3) Activation in the right homologues of the first pattern in Broca’s and Wernicke’s territories (1 control, 8 patients). Normal controls were predominantly in group 1; Patients fell into the three groups but distribution was different than controls, and were over represented in groups 2 and 3 (P<0.0004). Conclusions: Collaborative efforts to collect common measures using advanced imaging are feasible and practicable. In addition to the expected atypical right dominant activation pattern, a subset activation pattern was identified that involved different weighting within the distributed left hemisphere language processing network favoring frontal regions. This pattern, also seen in a minority of controls, may represent an alternative, less mature, or less efficient cognitive strategy for task primarily utilized by the LRE group.