fMRI analysis parameters affect the concordance with TMS in noninvasive speech mapping

BackgroundSpeech cortical mapping (SCM) conducted with widely available functional MRI (fMRI) can yield divergent results compared to the more commonly used navigated TMS (nTMS). The impact of specific fMRI task paradigms and preprocessing choices on reaching similarity with nTMS has not been explored before. ObjectiveTo test how the fMRI experimental task and spatial smoothing of the data compare with nTMS-based results, to subsequently increase the reliability of object naming fMRI for SCM. MethodsThirteen healthy, right-handed Finnish speakers underwent an nTMS-based SCM experiment in which the left hemisphere was stimulated while the subjects overtly named common visually presented stimuli. Standard as well as magnetoencephalography-informed picture-to-TMS intervals were applied. The same participants completed fMRI with overt naming, silent naming, and observation tasks on the same stimuli, analyzed with 0-, 3-, and 6-mm spatial smoothing. nTMS-based error and non-error sites were converted to volumetric density maps, and error-specific maps were derived by subtracting non-error from error density. Spatial similarity between binarized fMRI maps and nTMS maps was quantified using Jaccard index. Within-session fMRI reliability was estimated with voxel- and subjectwise concordance correlation coefficients across two separate runs conducted on the same day. ResultsSimilarity between fMRI and nTMS maps was overall low but depended significantly on data smoothing. Within subjects, mean error-specific Jaccard index was 0.036, with most individuals showing maximal similarity at 6 mm of smoothing. The fMRI task resulting in highest similarity with the nTMS map varied across participants, but at the group level, silent naming with 6-mm smoothing yielded the best correspondence. In general, within-session fMRI reliability increased with greater smoothing. ConclusionThe amount of applied fMRI data smoothing shapes the agreement of fMRI and nTMS maps during SCM. Silent naming fMRI combined with 6-mm data smoothing yielded the highest overlap with nTMS maps, yet the effect of the experimental task was statistically non-significant and the absolute similarity of the maps remained low. These results underline the different views to brain functions provided by direct perturbation of neural functions vs. blood-oxygenation based fMRI, and offer practical guidance when combining fMRI with nTMS in noninvasive speech cortical mapping. HighlightsO_LICorrespondence of fMRI and TMS speech cortical mapping results varied across individuals C_LIO_LIConcordance between the methods was generally low and depended on the fMRI data smoothing C_LIO_LISilent naming task in fMRI, combined with 6-mm data smoothing, yielded highest similarity to nTMS C_LIO_LIWithin-session fMRI reliability increased with greater smoothing C_LI