Synergistic and redundant information dynamics are modulated by Alzheimer's disease and cognitive impairment

Abstract

BackgroundThe early diagnosis of Alzheimers disease (AD), a cause of progressive cognitive decline, remains challenging. Recent information-theoretic advances allow brain dynamics to be quantified in terms of how regions share and combine information. Integrated Information Decomposition ({Phi}ID) separates redundant (the same content present in multiple regions) from synergistic information (new content that emerges only when regions are considered together). Such information-dynamic measures may provide biomarkers relevant to AD risk and progression. MethodsHere we applied integrated information decomposition ({Phi}ID) to resting-state fMRI from the Alzheimers Disease Neuroimaging Initiative (ADNI), to test whether {Phi}ID measures are diagnostically sensitive and track cognition along the AD spectrum. For each region, we computed total synergy and redundancy and compared values across cognitively normal (CN), mild cognitive impairment (MCI), and AD groups. ResultsCompared to CN, AD patients showed a striking synergy reduction across the entire brain, in concert with widespread redundancy increases, particularly in the executive and default mode networks. Transitions from CN to AD included an intermediate MCI decrease in redundancy, possibly reflecting early disease compensation strategies. This AD informational shift from complex higher level information processing to more robust inefficient forms likely reflects a cognitive shift to simpler, less integrative cognitive processes. Indeed, when re-analysing the data according to a standard cognitive clinical test (the Montreal Cognitive Assessment), we found a synergy-redundancy shift in high versus low performers broadly very similar to the CN to AD shift. ConclusionAD shows a clear information-processing signature: reduced global synergy and increased redundancy, especially in the executive control network. These striking results provide powerful insights into the widespread information processing reconfiguration that occurs in AD, with clear changes already emerging at the earlier MCI stage. Further, these results provide a novel route to support early diagnosis and stratification.