Resting State Functional MR Imaging of Language Function

Abstract

Resting state functional MR imaging methods can provide localization of the language system; however, presurgical functional localization of the language system with task-based functional MR imaging is the current standard of care before resection of brain tumors. These methods provide similar results and comparing them could be helpful for presurgical planning. We combine information from 3 data resources to provide quantitative information on the components of the language system. Tables and figures compare anatomic information, localization information from resting state fMR imaging, and activation patterns in different components of the language system expected from commonly used task fMR imaging experiments.

Keywords: Convolutional neural network; Deep learning; FreeSurfer; Language; Neurosynth; Presurgical planning; Resting state fMRI; Task fMRI.

Fig. 1.

Maps of the probability of RSNs (rainbow) and maps of correlations drawn from meta-analyses of task activations (gray-scaled) are superimposed on the Montreal Neurologic Institute atlas of 152 normal T1-weighted MR imaging scans (MNI152). A 3D CNN estimated the probability that voxels from resting state fMRI of normal subjects belonged to each of 9 classes of RSNs. Neurosynth provides correlations associated with tests of significance on meta-analyses of language-relevant topical terms. (A) Only Neurosynth correlations. (B–J) Fusions of RSNs and Neurosynth correlations in decreasing order of similarity.

Fig. 2.

Heatmaps indicate the Jaccard indices between anatomy identified by FreeSurfer, task functional regions identified by Neurosynth and RSNs identified by a deep 3D CNN. Between 2 binarized images, the Jaccard index is a segmentation measure that is, the intersection of voxels divided by the union of voxels. Jaccard indices have been multiplied by 100 and rounded to integers for clarity of display. (A) Comparing RSNs with Neurosynth. (B) Comparing FreeSurfer with RSNs. (C) Comparing Neurosynth with FreeSurfer. Thickened lines indicate the grouping of frontal, parietal, and temporal lobar anatomies. PMN, parietal memory network.

Fig. 3.

Heatmaps indicate the BF1 contour matching score between anatomy revealed by FreeSurfer, task functional regions revealed by Neurosynth and RSNs revealed by a deep 3D CNN. BF1 is a semantic segmentation measure that emphasizes the similarity of boundaries between 2 binarized images, complementing the description of overlapping voxels provided by the Jaccard index. BF1 indices have been multiplied by 100 and rounded to integers for clarity of display. (A) Comparing RSNs with Neurosynth. (B) Comparing FreeSurfer with RSNs. (C) Comparing Neurosynth with FreeSurfer. Thickened lines indicate the grouping of frontal, parietal, and temporal lobar anatomies. PMN, parietal memory network.