Abstract

Overt speech production in functional magnetic resonance imaging (fMRI) studies is often associated with imaging artifacts, attributable to both movement and susceptibility. Various image-processing methods have been proposed to remove these artifacts from the data but none of these methods has been shown to work with continuous overt speech, at least over periods greater than 3 s. In this study natural, continuous, overt sentence production was evaluated in normal volunteers using both arterial spin labeling (ASL) and conventional echoplanar blood oxygenation level-dependent (BOLD) imaging sequences on the same 1.5-T scanner. We found a high congruency between activation results obtained with ASL and the de facto gold standard in overt language production imaging, positron emission tomography (PET). No task-related artifacts were found in the ASL study. However, the BOLD data showed artifacts that appeared as large bilateral false-positive temporopolar activations; percent signal change estimated in these regions showed signal increases and temporal dynamics that were incongruent with typical BOLD activations. These artifacts were not distributed uniformly, but were aligned at the frontotemporal base, close to the oropharynx. The calculated head movement parameters for overt speech blocks were within the range of the rest blocks, indicating that head movement is unlikely the reason for the artifact. We conclude that ASL is not influenced by overt speech artifacts, whereas BOLD showed significant susceptibility artifacts, especially in the opercular and insular regions, where activation would be expected. ASL may prove to be the method of choice for fMRI investigations of continuous overt speech.