Impaired brain intrinsic connectivity in long COVID during cognitive exertion revealed by independent component analysis

Abstract Cognitive dysfunction is a symptom of Long COVID. To characterize functional connectivity changes that may contribute to cognitive dysfunction in Long COVID (LCov), two consecutive 450 s functional magnetic resonance imaging (fMRI) scans (Runs 1 and 2) were acquired on a 7 Tesla MRI scanner. During both, the Stroop colour-word task engaged intrinsic brain networks for conflict detection, conflict resolution and response execution. In this exploratory study we acquired data from 19 LCov and 16 healthy control (HC) participants. The aggregate dataset was subjected to independent component analysis (ICA) for each run to isolate 15 components with distinct spatial and temporal signatures. For each component we tested (1) for differences between LCov and HC (inter-network) connectivity to the rest of the brain and (2) for correlation of LCov connectivity with illness duration. Stroop response times (RTs) were slower in LCov than in HC in both Runs ( p = 0.001, 0.003). Each ICA component occupied the hubs of a known intrinsic network. LCov had different inter-network connectivity to HC for Salience, Language, Central Executive, Sensorimotor and Visual networks. LCov deficits in Salience inter-network connectivity in Run 2 were deeper and more widespread than in Run 1. In contrast, Run 2 connectivity was greater in LCov for the angular gyrus which integrates visual, motor and language inputs and responses. With longer illness duration, LCov connectivity weakened to critical networks but showed compensatory effects in Lingual gyri. Slower Stroop RTs in both Runs, and Salience, Language and Central Executive inter-network connectivity deficits, and illness duration dependence, support cognitive impairment in LCov with some compensatory connectivity increases.