Diffusion Tensor Imaging Measures are Associated with Speech Perception Ability in Young Cochlear Implant Candidates
Other conference paper


Full Text

Other information
AbstractIntroduction: A method to accurately predict spoken language after cochlear implantation would be an
important tool that could lead to customized therapeutic habilitation for children at risk for poor spoken
language development. Our research group has created predictive models of language development
with machine learning approach and neuroanatomy measures. Our models had superior predictive
ability of high versus low spoken language improvement compared to traditional predictors such as age
at implant. We report the first phase of our diffusion tensor imaging (DTI) study in which DTI measures
are correlated with pre-implant speech perception performance. DTI takes advantage of differential
diffusion of water molecules along tissues. It is used primarily to study axonal tracts because diffusion
occurs more readily and is directionally-dependent (anisotropic) along axons. We hypothesize that
neural connections measurable by DTI may be affected by hearing deprivation and that these
differences may be predictive of speech perception ability.
Methods: This study included 35 cochlear implant (CI) candidates (16.74±9.44 months, 17 female); and
35 age and gender-matched normal hearing children (17.55±9.53 m, 18 female) drawn from an NIH
database. The primary speech perception measure was the Speech Recognition Index in quiet (SRI-Q).
Fractional anisotropy (FA) maps were estimated for each subject and warped into a children-specific
brain atlas. The multi-voxel pattern similarity was computed over the entire FA map for each subject.
Whole-brain tractography was conducted and the FA values along the tracts connecting any pairs of the
predefined nodes were extracted to construct a connectivity matrix for each subject. Statistical
Analyses. Between- and within-group multi-voxel pattern similarity measures in the normal hearing
group were compared for local features (local white fiber integrity). Between- and within-group
similarities of the whole brain connectivity matrices were compared for global features. The betweengroup similarities of local and global measures in the CI group were correlated with speech perception.
Results: Regions primarily located in left superior longitudinal fasciculus show significant group
differences in multi-voxel pattern similarity between hearing and CI groups. The global similarities of
connectivity matrices were also different between hearing and CI groups (P < 0.001). And the similarities
of connectivity matrices between hearing and CI groups are significantly correlated with SRI-Q at
baseline (P =0.003).
Conclusion: DTI demonstrates abnormalities of local and global structural connectivity of CI candidate
brains. The global but not local connectivity patterns are associated with speech perception ability: the
more similar the patterns of axonal connectivity of the CI candidates to normal hearing children, the
higher their speech perception ability. The DTI-derived neural connectivity measures may be useful
neural makers for predicting language development. If so, our models may aid in understanding the
dose and type of therapy needed to maximize each child’s spoken language, the long term goal of our
research.
All Author(s) ListGeng X., Feng G., Ryan M.R., Wong P.C.M., Young N.M.
Name of ConferenceCI 2019 Pediatrics 16th Symposium on Cochlear Implants in Children
Start Date of Conference10/07/2019
End Date of Conference13/07/2019
Place of ConferenceHollywood
Country/Region of ConferenceUnited States of America
Year2019
LanguagesEnglish-United States

Last updated on 2020-15-06 at 15:32