Dynamic causal modelling of effective connectivity from fMRI: Are results reproducible and sensitive to Parkinson's disease and its treatment?

Abstract

Dynamic causal modelling (DCM) of functional magnetic resonance imaging (fMRI) data offers new insights into the pathophysiology of neurological disease and mechanisms of effective therapies. Current applications can be used both to identify the most likely functional brain network underlying observed data and estimate the networks' connectivity parameters. We examined the reproducibility of DCM in healthy subjects (young 18–48 years, n = 27; old 50–80 years, n = 15) in the context of action selection. We then examined the effects of Parkinson's disease (50–78 years, Hoehn and Yahr stage 1–2.5, n = 16) and dopaminergic therapy. Forty-eight models were compared, for each of 90 sessions from 58 subjects. Model-evidences clustered according to sets of structurally similar models, with high correlations over two sessions in healthy older subjects. The same model was identified as most likely in healthy controls on both sessions and in medicated patients. In this most likely network model, the selection of action was associated with enhanced coupling between prefrontal cortex and the pre-supplementary motor area. However, the parameters for intrinsic connectivity and contextual modulation in this model were poorly correlated across sessions. A different model was identified in patients with Parkinson's disease after medication withdrawal. In “off” patients, action selection was associated with enhanced connectivity from prefrontal to lateral premotor cortex. This accords with independent evidence of a dopamine-dependent functional disconnection of the SMA in Parkinson's disease. Together, these results suggest that DCM model selection is robust and sensitive enough to study clinical populations and their pharmacological treatment. For critical inferences, model selection may be sufficient. However, caution is required when comparing groups or drug effects in terms of the connectivity parameter estimates, if there are significant posterior covariances among parameters.

The acute brain response to levodopa heralds dyskinesias in Parkinson disease

“…Interestingly, activation of preSMA has been linked to internally generated movements and the intention to act,35 including the generation of involuntary actions 37. The connectivity of this region to motor regions is abnormal in PD38 and modulated by dopamine 39. Furthermore, preSMA is involved in decreasing motor threshold during the speed–accuracy tradeoff 40.…”

Depression alters “top-down” visual attention: A dynamic causal modeling comparison between depressed and healthy subjects

“…Given the nature of the brain mechanism modelled here (basic level visuo-attentional process) and the homogeneity of the subjects in each group, a FFX approach is preferred here both at the family and model inference level. Indeed with a FFX scheme, one assumes that every subject (in a group) uses the same model architecture, an assumption commonly accepted (Kumar et al, 2007;Stephan et al, 2007a,b;Acs and Greenlee, 2008;Rowe et al, 2010) and warranted when studying a basic physiological mechanism that is unlikely to vary across subjects . The results of the RFX analysis, both at family and model level, are available in the supplementary material section for comparison.…”

“…In fact Bayesian estimation and averaging account for the covariance between the parameters. A strong or consistent covariance between a pair of parameters across subjects could lead to less reliable parameter estimates when looked at individually, leading to a reduced sensitivity (Rowe et al, 2010). Thus the estimated posterior of a single parameter could well be not significantly different from zero on average but the connection modelled still necessary from a model selection point of view.…”

Dynamic Causal Modeling of Preclinical Autosomal-Dominant Alzheimer’s Disease

“…The majority of clinical applications of DCM have been made using fMRI and range from studies of aphasia [14], autism [15,16], and major depression [17,18] to Parkinson's disease [19] and schizophrenia [20][21][22]. Further applications in psychiatry are reviewed in Yu et al [23].…”

“…In this fMRI study of speech processing, connectivity estimates from a DCM of thalamo-temporal regions provided discrimination between moderately aphasic and healthy control groups that was better than that achievable using conventional activation-based and correlationbased methods. Second, DCM was used to study changes in brain connectivity due to an action selection task in two groups of subjects: control subjects and subjects with Parkinson's disease who were undergoing dopaminergic therapy [19]. The optimal DCM was the same in both groups and showed modulation of coupling between prefrontal cortex and the pre-supplementary motor area (SMA).…”