Concurrent neuroimaging and neurostimulation reveals a causal role for dlPFC in coding of task-relevant information

Abstract

The way in which the brain prioritises processing of information relevant for our current goals is widely contested. Many studies implicate the dorsolateral prefrontal cortex (dlPFC), and propose that it drives brain-wide focus by biasing processing in favour of relevant information. An alternative, however, is that dlPFC is involved in the inhibition of irrelevant information. Here, we address this longstanding debate using the inferentially powerful approach of applying transcranial magnetic stimulation during functional magnetic resonance imaging (concurrent TMS-fMRI) and testing for changes in information coding using multivariate pattern analysis (MVPA). We ask whether dlPFC plays a causal role in prioritising information processing, and whether this is through selection of relevant information or inhibition of irrelevant information. Participants attended to one object feature whilst ignoring another feature of the same object. We reasoned that, if dlPFC is necessary for selection, active (disruptive) TMS should decrease coding of attended information compared to the low intensity (control) condition. Conversely, if right dlPFC is crucial for inhibition, active TMS should increase coding of irrelevant information relative to the control condition. The resultsshowed that active TMS decreased coding of relevant information throughout the frontoparietal multiple demand regions, and that this impact was significantly stronger than the effect of TMS on irrelevant information coding, which was not statistically detectable. These data provide causal evidence for a specific role of dlPFC in supporting the representation of task-relevant information and demonstrate the crucial insights into high level cognitive-neural mechanisms possible with the combination of TMS-fMRI and MVPA.

Selective attention and decision-making have separable neural bases in space and time

“…In addition, Woolgar and colleagues (2015) found stronger visual object coding in the MD cortex for cued (i.e., attended, task-relevant) objects relative to equivalent distractor objects, and Jackson and colleagues (2016) reported stronger MD coding of cued compared to equivalent distractor features of the same object. Moreover, MD coding of cued information can be selectively impaired by non-invasive stimulation the right dorsolateral prefrontal cortex, one of the nodes in the MD network (Jackson et al, 2020). In parallel, recent magnetoencephalography (MEG) studies have shown the coding of a cued feature at a cued location is sustained over time, while information coding for distractor information is not sustained (Battistoni et al, 2018; Goddard et al, 2019).…”

Neural coding of visual objects rapidly reconfigures to reflect sub-trial shifts in attentional focus

“…For example, both rule and relevant stimulus information can typically be decoded from MD regions in human fMRI (Jackson et al, 2016;Woolgar, Afshar, et al, 2015;Woolgar & Zopf, 2017) and from frontal cortex in non-human primate single-unit recordings (Everling et al, 2006;Stokes et al, 2013). Disrupting prefrontal function causes reduction in task-relevant information coding (Jackson et al, 2021), and incorrect rule or stimulus information coding predicts incorrect behavioural responses (Woolgar et al, 2019). Moreover, the structure of frontal stimulus information predicts subsequent occipital stimulus information as attentional selection of relevant features emerges (Goddard et al, 2019).…”

“…Moreover, MD coding for task-relevant stimuli is enhanced when stimuli are more difficult to discriminate (Woolgar, Hampshire, et al, 2011; and changes to prioritise information that is at the focus of attention (Jackson & Woolgar, 2018;. Activity in at least one MD region appears to be causal for facilitating taskrelevant information processing (Jackson et al, 2021). This, in turn, may provide a source of bias to more specialised brain regions, for example through task-dependent connectivity (Cole et al, 2013; see for example Baldauf & Desimone, 2014).…”

Now you see it, now you don't: optimal parameters for interslice stimulation in concurrent TMS-fMRI

“…One approach is to remove the affected slices or volumes via temporal interpolation (Fig. 1a) [6, 7, 15]. This can be controlled to result in minimal information loss [16] but is laborious and requires careful assessment to ensure that all affected slices have been removed.…”

“…However, offline approaches are not suited for all paradigms, for example, trial-by-trial investigations or where the timing of TMS delivery is critical, and are limited by the duration of the induced TMS effect. Thus, several studies have adopted concurrent methods, for example, to investigate causal influences of higher order brain areas on specialised cortices [2][3][4][5][6][7]. Recently there has been a resurgence of interest in concurrent TMS-fMRI, for example combining with modern analytical techniques for causal insights into information processing [8,9].…”