Abstracts

Rationale: Overt speech is not typically investigated with functional Magnetic Resonance Imaging (fMRI) due to the risk of in-scanner motion reducing data quality. As such covert language paradigms without overt speech responses are commonly used to assess expressive language dominance in epilepsy patients considered for neurosurgical intervention. The extent to which covert responses are a valid model for assessing the neural correlates of overt speech has not been systematically investigated. Overt paradigms also provide the advantage of online performance monitoring, which can help clinicians interpret fMRI findings. Despite this, there have been no systematic investigations to assess the feasibility of imaging overt speech in patient populations when capitalizing on new neuroimaging technology and artifact repair software.Methods: Twenty pre-surgical pediatric epilepsy patients (6 males, 14 females, 7-17 years) performed covert and overt verb generation (VG) tasks within the 1.5T MRI scanner. Speech responses were recorded using a sensitive head mounted microphone (MR-Confon Ltd). Data were analyzed with SPM8 software using a blocked design with a baseline condition of listening to white noise. The SPM ArtRepair toolbox was used to assess data quality and to repair movement contaminated data. Data quality was reviewed blindly by 3 independent investigators. Standard t-contrasts investigated differences in neural substrates between conditions (random effects analysis). Laterality indices(LI) were calculated in predefined regions of interest using the SPM LI toolbox, to investigate the effect of response modality on measures of language dominance used in clinic.Results: Overt speech was associated with increased in-scanner motion (p<0.005) but also increased peak activation within the language network (p=0.001). Prior to repairs 33% of overt VG data were rated as good quality compared to 21% of covert VG data. Data quality could be improved in 45% of high artifact cases. Overt verb generation increased recruitment of the non-dominant IFG, posterior STS, SMA and pre-motor cortex relative to covert VG. Patients with typical language dominance (N= 12) demonstrated a left-lateralized LI in Broca s area during covert VG (LI = 0.78) but a more bilateral LI during overt VG (LI = 0.26, p<0.05). Furthermore, patients were more frequently categorized as having atypical language dominance based on overt VG data (53% of the sample) than covert VG data (40%).Conclusions: Language paradigms with overt speech responses produce robust activation within language regions relative to covert paradigms. Although overt responding is associated with increased in scanner movement, in the majority of cases this artifact can be rectified using post-processing. As such, this study suggests overt investigations of expressive language are both viable in pediatric patient populations and provide a valid assessment of the neural substrates engaged during real-life speech. This is particularly relevant in the pre-surgical setting, when the margin for error is low.