Short- and long-term reliability of language fMRI

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Neurology

Background:

• High test-retest reliability of functional magnetic resonance imaging (fMRI) is essential for both basic research and clinical applications in mapping language functions.
• Previous studies have not systematically evaluated the reliability of fMRI for language tasks over different time intervals and across various brain regions.

Purpose of the Study:

• To systematically assess the short- and long-term test-retest reliability of fMRI in healthy subjects performing an overt picture naming task.
• To investigate whether reliability differs between speech-related motor areas (M1, IFG) and other language areas (STG).
• To compare the reliability of fMRI measures over short (2-6 days) versus long (21-34 days) intersession intervals.

Main Methods:

• 16 healthy, right-handed subjects underwent three fMRI sessions using a sparse-sampling protocol.
• Participants performed an overt picture naming task (50 images per session).
• Reliability was quantified using Euclidean distances between activation maxima and Centers of Gravity (CoGs), overlap volumes, and voxel-wise intraclass correlation coefficients (ICCs) in primary motor cortex (M1), superior temporal gyrus (STG), and inferior frontal gyrus (IFG).

Main Results:

• Activation centers (maxima and CoGs) demonstrated high reliability, irrespective of the time interval, region of interest (ROI), or hemisphere.
• The extent of activation showed lower reliability, with overlap volumes ranging from 24% (IFG) to 56% (STG), and higher overlap in the left hemisphere.
• While mean ICCs indicated poor to moderate reliability, highly reliable voxels (ICC > 0.75) were identified across all ROIs, with voxel-wise reliability influenced by the intersession interval.

Conclusions:

• Functional magnetic resonance imaging (fMRI) reliably identifies activation centers during picture naming tasks, even over extended periods.
• The extent of activation is less reliable over time, with significant variations across ROIs and hemispheres.
• Despite ROI-dependent variations in activation extent, the study confirms the feasibility of using overt picture naming paradigms for reliable fMRI-based language mapping.