Integration of multimodal MRI data via PCA to explain language performance.

Area of Science:

• Neuroimaging
• Epilepsy Research
• Cognitive Neuroscience

Background:

• Single-modality neuroimaging has limitations in understanding brain-disease relationships.
• Limited knowledge exists on how various MRI measures correlate with language function in neurological diseases.
• Temporal lobe epilepsy (TLE) often involves language dysfunction, but its neuroimaging correlates require further investigation.

Purpose of the Study:

• To explore relationships between regional cortical thickness, gray-white matter contrast (GWMC), and white matter diffusivity (mean diffusivity [MD] and fractional anisotropy [FA]).
• To determine the predictive power of these MRI measures for language function (vocabulary, naming, fluency) in TLE patients and healthy controls.
• To assess the combined utility of multimodal MRI in predicting language impairment.

Main Methods:

• Collected T1- and diffusion-weighted MRI data from 56 healthy controls and 52 TLE patients.
• Focused on frontotemporal regions implicated in language.
• Reduced MRI data to principal components (PCs) and analyzed correlations and predictive abilities for language tasks.

Main Results:

• Significant positive associations found between PCs for cortical thickness, GWMC, and FA, partially mediated by total brain volume.
• Reduced FA correlated with increased MD, and reduced FA was linked to visual naming deficits, while increased MD was linked to auditory naming deficits.
• Inclusion of FA and MD PCs significantly improved the sensitivity and specificity of models predicting language impairment.

Conclusions:

• Quantitative MRI measures from T1 and diffusion-weighted scans are intercorrelated, suggesting a simultaneous pathological process affecting cortical and subcortical structures in TLE.
• Diffusion measures (FA and MD PCs), alongside hippocampal volume, enhance prediction of naming impairment in TLE.
• Combining multimodal imaging measures is crucial for better prediction of language performance in TLE and potentially other conditions with language impairments.