Slow-4 cerebellar activity tracks MMSE language-related task performance across the Alzheimer's disease spectrum: a...

Performance on MMSE language-related tasks may decline across the Alzheimer's disease (AD) spectrum and has clinical relevance for characterizing cognitive change. However, because the MMSE language subscore is a brief composite screening measure rather than a domain-specific language battery, its neural correlates should be interpreted cautiously. The intrinsic neural activity patterns associated with MMSE language-related task performance across the AD continuum remain insufficiently characterized. Most prior studies have examined isolated functional metrics or focused on global cognition, leaving unclear whether multidimensional spontaneous activity measures and their frequency-specific properties converge on common neural substrates associated with this composite behavioral measure.

We analyzed resting-state fMRI data from 175 ADNI participants (32 AD, 90 MCI, 53 CN), computing four complementary amplitude-based metrics-ALFF, fALFF, PerAF, and wavelet-ALFF-with wavelet-ALFF further decomposed into slow-4 (0.027-0.073 Hz) and slow-5 (0.01-0.027 Hz) bands. To examine functional alterations associated with MMSE language-related task performance, participants were grouped according to MMSE language sub-scores into intact performance, mild language-related impairment, and greater reduction in performance. Voxel-wise ANCOVA, partial correlation, and multiple linear regression analyses were performed to evaluate associations between resting-state metrics and MMSE language-related task performance, modeled both categorically and continuously, while controlling for age, sex, and education. Multiple comparisons were corrected using Gaussian Random Field theory.

Across diagnostic groups, amplitude-based metrics revealed frequency-dependent alterations: slow-5 band changes predominated in temporal regions, while slow-4 band alterations characterized parietal areas, particularly the right angular gyrus (127 voxels, peak F = 24.44). When stratified by MMSE language-related task performance, the paravermal (predominantly right) cerebellar lobule VIII emerged as the only region consistently identified across multiple metrics, including ALFF, PerAF, wavelet-ALFF, and across analytic approaches. Critically, the cerebellar lobule VIII finding showed its most consistent cross-analytic support in the slow-4 band. In ROI-based partial-correlation analyses controlling for age, sex, and education, higher slow-4 wavelet-ALFF in this region was negatively associated with MMSE language-related task performance, and this association remained significant after additional adjustment for global disease severity (CDR-SB; partial r = -0.31, p < 0.001) and for non-language MMSE performance (partial r = -0.31, p < 0.001). The negative direction-whereby greater activity in this cerebellar cluster accompanied poorer performance-was consistent across conventional broadband, slow-4, and slow-5 analyses.

These findings suggest that right cerebellar lobule VIII activity is meaningfully associated with performance on MMSE language-related tasks across the AD spectrum, particularly within the slow-4 frequency range. Importantly, because the MMSE language subscore includes naming, repetition, reading, writing, command execution, and constructional praxis, the observed associations should not be interpreted as evidence for a purely language-specific cerebellar mechanism. Rather, they may reflect a combination of linguistic, executive, visuospatial, motor-planning, and sensorimotor processes required by MMSE language-related tasks. More broadly, the present study highlights the value of frequency-resolved spontaneous activity analyses for identifying clinically relevant brain-behavior associations in AD and suggests that cerebellar circuits warrant further investigation in relation to composite cognitive and language-related task performance.