Annual Meeting Abstracts: View

  • (Abst. 3.229), 2017
  • Novel data-driven method for language dominance derived from resting-state and language task fMRI functional connectivity in Epilepsy patients
  • Authors: Xiaozhen You, Children's National Medical Center; Madison Berl, Children's National Medical Center; Charles Lynch, Georgetown University; Leigh N. Sepeta, Children’s National Medical Center, George Washington University School of Medicine and Health Sciences, Washington, DC, United States.; Chandan Vaidya, Georgetown University; and William D. Gaillard, Children’s National Medical Center, George Washington University School of Medicine and Health Sciences, Washington, DC, United States.
  • Content:

    Rationale: Presurgical language mapping is crucial to help evaluate the risk of postoperative deficits for epilepsy surgery. While fMRI provides a non-invasive method for this, task compliance is difficult for very young or intellectually impaired patients to obtain reliable activations. Resting state fMRI(rs-fMRI), which does not require task compliance might be a viable solution. Efforts to use rs-fMRI to identify language laterality are few, most are at group level, or a priori regions of interest (ROI) dependent, or require subjective evaluation of spatial components, and clinical validation is lacking, especially at an individual basis. We applied previously developed a data-driven method to identify language network laterality through functional connectivity (FC) analysis, which validated in the Human Connectome Project sample at both language task and resting state, in clinical pediatric epilepsy patients, and examined the correspondence between language task activation and both rest and task derived FC measures. Methods: We assessed the novel FC metric during 5 minutes age-adjusted language task (Auditory Description Decision Task) and a 5 minutes resting state scan in children with epilepsy (n=43,4 with tumor; mean age=10.1; SD=2.1). fMRI data went through similar pipeline as Human connectome projects. We applied denoising with aCompCor strategy and bandpass filter (0.01-0.1Hz) in CONN toolbox plus “scrubbing” high motion volumes and regressed out task structure for task data. Freesurfer analysis was done on patients’ structural T1 to map cortical ribbon onto each individual’s 32K_fs_LR mid-thickness surfaces (via Connectome Workbench).For each vertex within gray matter, we counted how many vertex in the language target mask ( were connected to it both Ipsilaterally (Intra) and Contralaterally (Inter). Then the FC-HC for each vertex was calculated as Intra-Inter at different r threshold (determined from a fixed edge density, i.e.,10% and 20% of the possible connections within target mask with r varies by subjects, as well as fixed rs:0.1-0.8 with step 0.1). A map of vertex-wise FC-HC (range 0-1, the percentage of all possible connections to the mask from that vertex) was then generated to indicate each individual's laterality.We extracted top 5,10, and 15% FC-HC or task activation for Broca’s (IFG) and Wernicke’s (BA21,22,39,40) to categorize the language dominance similarly to classic language fMRI activation laterality index (LI): (L-R)/(L+R) with >0.2 being left lateralized, <-0.2 right, and otherwise bilateral. We then did an exploratory search for categorical agreement between activation and FC-HC measures (at all r thresholds).  Results: Patients as a group showed left dominant language activation, similar left dominant task FC-HC but a more bilateral rest FC-HC, with paired t test showed increased FC-HC from rest to task on the left key language area (Fig.1).  At individual level, we found 75% patients at task and 67% at rest agreed on Broca with task activation, while 70% agreed at task and 60% at rest on Wernicke. Low agreement might result from less reliable task activation used as gold standard. Conclusions: We demonstrated the potential of using our novel data-driven method to parse the language network and establish reliable language laterality in task-free fMRI as well as task fMRI.  This may help expand the use of clinical fMRI because a successful study would no longer depend upon task performance of the epilepsy patient, opening its utility up to younger and more impaired populations.  Funding: Funded by R01 NS44280 to W.D.G., M01RR020359 K23NS065121-01A2 to M.B
  • Figures:
  • Figure 1