Predicting Auditory Skill Outcomes After Pediatric Cochlear Implantation Using Preoperative Brain Imaging

Purpose: Our study used preoperative neuroanatomical features to predict auditory development in Chinese-learning children with cochlear implants (CIs). Method: T1-weighted whole-brain magnetic resonance imaging (MRI) scans were obtained from 17 Chinese-learning pediatric CI candidates (12 females and five males, age at MRI = 23.0 ± 15.0 months). Voxel-based morphometry was applied to examine the children’s whole-brain structure. Machine learning was employed using neuroanatomical features to predict children’s auditory skills up to 24 months after CI. The whole-brain neural model and auditory/visual cortex neural model were compared with a nonneural model using gender, age at CI activation, and preoperative residual hearing as predictors. Model performance was quantified using the mean square error (MSE) between predicted values and observations. Results: The model with preoperative neuroanatomical features showed a sig­nificantly smaller MSE than the nonneural model in predicting auditory skills in children with CIs. Specifically, the auditory-related area played an important role in predicting post-CI outcomes. Conclusions: The preoperative neuroanatomical features outperformed the non- neural features in predicting auditory skills in children with CIs. These results indicate that neural structure holds the potential to serve as an objective and effective feature for predicting post-CI outcomes. Supplemental Material: https://doi.org/10.23641/asha.28012046