Ketamine infusion in healthy men increases blood-oxygen-level dependent signals across the cortex and decreases them in the subgenual anterior cingulate cortex, but the decrease is largely due to physiological noise, especially cardiac pulsatility, rather than neural activity. Modeling the pharmacological MRI response with a single time course misses the full range of neural dynamics; using simultaneously recorded electroencephalography power time series reveals distinct temporal responses to ketamine, though no EEG band correlated with the subgenual anterior cingulate cortex decrease.
The validity of fMRI functional connectivity as a drug biomarker was tested by comparing seven preprocessing pipelines and by simultaneously measuring EEG and fMRI in a placebo-controlled, three-way crossover study with ketamine and midazolam. Independent components analysis (ICA)-denoising produced stronger reductions in connectivity after ketamine and weaker increases after midazolam than pipelines using physiological noise modelling or averaged signals from cerebrospinal fluid or white matter. This indicates that pipeline decisions should match a drug's unique noise structure; when unknown, extensive ICA denoising may sacrifice some signal but increase confidence in remaining results. No significant relationship was found between changes in electrophysiological and hemodynamic correlation structures, cautioning against cross-modal comparisons of pharmacologically-modulated functional connectivity.