The study analyzed previously acquired MRI and clinical data from a single tertiary hospital in Fuzhou, China. After 21 participants were excluded under prespecified head-motion and image-quality criteria, the final sample comprised 97 adults: 38 healthy controls, 29 patients with cirrhosis who had no previous overt hepatic encephalopathy and were classified as West Haven grade 0 at enrollment, and 30 patients in the operationally defined HE group. The mean age was 55.8 years, and 40 participants, or 41.2%, were women. The initial sample included 118 people, and the retrospective design did not establish whether imaging patterns preceded or followed changes in clinical status.

The operational HE group was heterogeneous. Of its 30 participants, 19 had current West Haven grade 1–2 HE and 11 had recovered from previous overt HE to grade 0 at enrollment. Seventeen members of the group had a previous overt episode, while 13 had current grade 1–2 HE without previous overt HE. The no-history group was defined by previous overt HE status and enrollment grade rather than by psychometric testing, so it could include patients with minimal HE. Grade 3–4 HE was excluded, and the authors noted that cirrhosis etiology, transjugular intrahepatic portosystemic shunt status, detailed prior HE burden, and HE-directed treatment exposure were not modeled separately.

The investigators calculated two voxel-wise ratios within a strict common gray-matter mask. ReHo-to-GMV divided standardized regional homogeneity by local gray matter volume, while ALFF-to-GMV divided standardized amplitude of low-frequency fluctuation by the same structural measure. Regional homogeneity describes local temporal concordance, whereas ALFF describes the magnitude of low-frequency fluctuations in the 0.01–0.08 Hz band. Separate three-group models adjusted for age, sex, education, and mean framewise displacement. Gaussian random-field cluster inference was the primary analysis, and permutation-based threshold-free cluster enhancement with 10,000 permutations was used as a non-parametric sensitivity analysis.

ReHo-to-GMV produced six corrected omnibus clusters and seven corrected directional clusters. The HE group had a higher ratio than controls in the right thalamus, while controls had higher ratios than the HE group in the anterior cingulate and medial frontal region, insular-putaminal-opercular areas, superior temporal cortex, and perirolandic-parietal regions. ALFF-to-GMV yielded three corrected omnibus clusters and ten directional clusters. The HE group had higher ratios in medial temporal and thalamus-adjacent regions, whereas controls had higher ratios in frontal, precuneus, middle temporal, supplementary motor, and perirolandic areas. Together, the two metrics produced 13 corrected clusters each, but their maps overlapped in only 31 voxels, all in the control-greater-than-HE direction, corresponding to a Dice coefficient of 0.0541.

Neither ReHo-to-GMV nor ALFF-to-GMV showed a corrected difference between the HE and no-history groups. An exploratory comparison within the HE group likewise found no corrected separation between 19 patients with current grade 1–2 HE and 11 who had recovered from previous overt HE, despite lower PHES, Mini-Mental State Examination and Montreal Cognitive Assessment scores in the current-HE subgroup in unadjusted comparisons. These null findings do not demonstrate equivalence: the groups were modest in size, clinically heterogeneous, and the analyses reused the same participants.

Sensitivity analyses broadly supported the main HE-versus-control patterns. Permutation-based threshold-free cluster enhancement confirmed effects for both ratios, including after additional joint correction across the three pairwise contrasts. Reconstructing the analysis mask from the final 97 participants reproduced all 26 primary clusters. Expanded motion adjustment retained all ReHo directional clusters, although two smaller ALFF control-greater-than-HE clusters were not retained, making those findings more sensitive to motion-model specification. Total-intracranial-volume adjustment also preserved strong or partial overlap for 10 of 13 primary clusters for each metric.

Voxel-wise models that directly adjusted for local gray matter volume indicated that the principal HE-versus-control effects were not solely attributable to denominator variation. For ReHo, 415 of 647 original control-greater-than-HE ratio voxels, or 64.1%, were retained in the same direction, with a Dice coefficient of 0.606. For ALFF, 94.7% of the original HE-greater-than-control voxels and all 500 original control-greater-than-HE voxels were retained with concordant direction, although the adjusted ALFF effects were more extensive. These checks do not make ratio maps and covariate-adjusted models interchangeable or establish independence from gray matter volume.

None of 48 exploratory associations between eight imaging-defined regions of interest and cognitive, ammonia, Child-Pugh or MELD measures survived false-discovery-rate correction. The study also had no external validation, longitudinal follow-up, or test-retest assessment, and the functional and gray-matter maps were smoothed with different kernels. Susceptibility-distortion correction and scrubbing were not performed, leaving particular concern in orbitofrontal, insular and medial temporal regions. The authors therefore interpret the results as distributed, metric-dependent group-level differences that require independent and longitudinal evaluation, rather than as a distinct HE-specific signature or an established patient-level diagnostic, staging, prognostic or monitoring tool.

Subject of Research: Voxel-wise functional-to-gray matter volume ratios in cirrhosis and operationally defined hepatic encephalopathy groups

Article Title: Voxel-wise functional-to-gray matter volume ratios reveal distinct brain patterns in cirrhosis stratified by hepatic encephalopathy history and clinical status

Article References: Yu, Z., Mao, M., Xue, Q., Hong, C., Huang, H., Shen, A., & Fu, L. (2026). Voxel-wise functional-to-gray matter volume ratios reveal distinct brain patterns in cirrhosis stratified by hepatic encephalopathy history and clinical status. Frontiers in Neurology, 17, Article 1969935. https://doi.org/10.3389/fneur.2026.1969935

Image Credits: AI Generated

DOI: 10.3389/fneur.2026.1969935

Keywords: cirrhosis, hepatic encephalopathy, resting-state fMRI, regional homogeneity, ALFF, gray matter volume, functional-to-volume ratios, brain imaging, clinical associations, Voxel-wise, functional-to-gray, matter