Dynamic reconfiguration of brain activity states in peritoneal dialysis patients: a Hidden Markov Model study

Cognitive decline is prevalent among peritoneal dialysis (PD) patients, but its neural mechanisms underlying resting-state latent brain state dynamics remain unclear. This study used Hidden Markov Model (HMM) to compare latent brain activity dynamics between PD patients and healthy controls (HC), and further explore their correlations with cognitive function. Resting‑state fMRI data from 50 PD patients and 46 matched HC were analyzed. Fractional occupancy, switching rate and transition probabilities derived from HMM were compared between groups, followed by correlation and logistic regression analyses. Six latent brain states were identified. PD patients had higher occupancy of States 1, 2, 4, 6 and lower occupancy of States 3, 5 (FDR-corrected P < 0.05). Global switching rate showed no group difference, yet PD patients displayed abnormal transitions among specific states. In a multivariable logistic model restricted to PD patients, lower State 3→5 transition probability was independently associated with mild cognitive impairment (adjusted OR = 0.60 per 0.1-unit increase, 95% CI 0.38–0.97, P = 0.037). State 2’s spatial pattern was preferentially expressed in PD patients. PD patients exhibit reconfigured latent brain activity states. Abnormal brain‑state transitions are related to cognitive impairment, offering potential neuroimaging insights for PD‑associated cognitive decline.