In this study, EEG-based microstate analysis provided insights based on the scalp topography, as shown by an illustrative example of the GFP of the active microstate dynamics in Fig. 6. At the group level, we found that microstates 1 and 3 were the most dominant (high GEV) across all conditions, and microstate 2 was found missing in the experts during 10-sec at the error epoch. The microstate 2 is most topographically similar to Brechet and colleagues’s (Bréchet et al., 2019) microstate A (see Figure S3, supplementary materials) that showed left-lateralized activity in the superior temporal gyrus (STG), the medial prefrontal cortex (MPFC) and the occipital gyri (OCG). Also, the microstate 2 is most topographically similar to Custo et al.'s (Custo et al., 2017) microstate A that represent the left middle and superior temporal lobe activity, which is postulated to be associated with the exploration of both object-related and space-related information (Karnath, 2001). Here, the missing microstate 2 in experts indicated lack of exploratory motor behavior in the error epoch. Then, during 10-sec at the start of the FLS complex task, microstate 1 accounted for the highest GEV for experts while microstate 3 accounted for the highest GEV for novices. Also, during 10-sec at the error epoch, microstate 3 accounted for the highest GEV for experts while microstate 1 accounted for the highest GEV for novices. The microstate 3 is topographically similar to Brechet and colleagues’s (Bréchet et al., 2019) microstate D (see Figure S3, supplementary materials) where the sources showed main activity bilaterally in the inferior frontal gyrus (IFG), dorsal anterior cingulate cortex (dACC), and superior parietal lobule (SPL)/intraparietal sulcus (IPS). Then, microstate 1 is most topographically similar to Brechet and colleagues’s (Bréchet et al., 2019) microstate C (see Figure S3, supplementary materials) where the sources showed sources located bilaterally in the lateral part of the parietal cortex including both the supramarginal gyrus (SMG) and angular gyrus (AG). Here, we postulate that the activation of SMG and AG during FLS complex task indicated ventral attention while the activation of SPL/IPS indicated the dorsal attention systems that are relevant in reorienting of visuospatial attention (Vossel et al., 2014). Here, ANOVA revealed a significant effect of skill level (expert, novice) on the proportion of the total time spent in the microstates during the 10-sec duration at the start of the FLS complex task and error epoch. Microstate 3 can also be topographically related to the canonical microstate D from the combined EEG-fMRI recording of resting state published by Britz et al. (Britz et al., 2010). Microstate D, published by Britz et al. (Britz et al., 2010), was shown by a behavioral manipulation study by Milz et al. (Milz et al., 2016) that reflected reflexive aspects of attention, focus switching, and reorientation (Michel and Koenig, 2018), which is necessary for error related switching of the mental state.
The HbO response and the image of the average changes in HbO absorption coefficient in the cortex during the 10-sec epoch are shown in Fig. 7. Statistical testing of the HbO hemodynamic response at the fNIRS channels identified underlying left postcentral gyrus and right superior frontal gyrus (SFG) – orbital part as significantly different between the experts and novices during 10-sec at the start of the FLS complex task while HbO hemodynamic response at the underlying left IFG – opercular part, left SFG – medial orbital, left postcentral gyrus, left STG, right SFG – medial orbital were significantly different between experts and novices during the error epoch. Here, the postcentral gyrus contains the primary somatosensory cortex and right SFG – orbital part contributes to the proactive control of the impulses (Hu et al., 2016) relevant in the performance of the FLS complex task that was different between experts and novices. Also, the activation of the left IFG – opercular part, left SFG – medial orbital can be related to higher cognitive functions (du Boisgueheneuc et al., 2006) while the activation of the right SFG – medial orbital can be related to the proactive control of the impulses (Hu et al., 2016) – both relevant in error processing that was different between experts and novices. Moreover, while the left postcentral gyrus is related to the motor action in the right-handed subjects, the STG can be related to the exploration (Karnath, 2001) during error processing that was different between experts and novices, viz., we found microstate 2 that is related to the left middle and superior temporal lobe (Custo et al., 2017) missing in the experts. We showed the fusion of information from simultaneously acquired EEG and fNIRS signals to provide a mechanistic insight into the changes in the brain state during FLS complex task as well as error perception and correction during FLS skill training.
Statistics on the transition probabilities between microstate classes at the group level showed that the novices mainly transitioned between microstate 3 to microstate 1 during the 10-sec at the start of the FLS complex task and from microstates 3,4 to microstate 1 in 10-sec of the error epoch where microstate 1 can be associated with posterior isoelectric point in the topographical map of the salience network (Santarnecchi et al., 2017) that is involved in attending to and responding to error (unexpected) stimuli. Here, microstate 4 is most topographically similar to Brechet and colleagues’s (Bréchet et al., 2019) microstate F (see Figure S3, supplementary materials) that showed strongest activity in the right MPFC. In contrast, the experts mainly transitioned from microstates 2,5 to microstate 3 during the 10-sec at the start of the FLS complex task and from microstate 5 to microstate 3, from microstate 6 to microstate 1 in the 10-sec of the error epoch. Here, microstate 5 is most topographically similar to Brechet and colleagues’s (Bréchet et al., 2019) microstate F (see Figure S3, supplementary materials) that showed bilateral activity in the MPFC. Then, microstate 6 is most topographically similar to Brechet and colleagues’s (Bréchet et al., 2019) microstate B (see Figure S3, supplementary materials) that showed main activity in OCG and in the medial part of the parietal cortex. Microstate 6 can be associated with spatial attention (Britz et al., 2010) so transition to microstate 1 of the salience network (Santarnecchi et al., 2017) highlights visual error awareness and salience processing of error (unexpected) stimuli in the experts. Moreover, the transition of microstates 1,4,5 to microstate 3 with high (> 0.6, see Figure s2 in Supplementary Materials) transition probabilities illustrate the learned reflexive aspects of attention, focus switching, and reorientation (Michel and Koenig, 2018) in the experts during error-related adjustments. Here, the microstate correlates of exploration-exploitation tradeoff, e.g., microstate 2 for motor exploration (Karnath, 2001) and microstate 3 with SPL activity for planning and guiding movement relevant in motor exploitation, are postulated to differentiate experts and novices (Phillips et al., 2011) during error-related adjustments that need further investigation in the future studies.
Our brain activation results aligned with numerous functional magnetic resonance imaging (fMRI) and fNIRS studies that have been published on skill learning (Roberts et al., 2006),(Ohuchida et al., 2009),(Leff et al., 2008c),(Wanzel et al., 2007),(Leff et al., 2007),(Gao et al., 2021a),(Leff et al., 2008b),(Khoe et al., 2020),(Gao et al., 2021b). Published fMRI studies have shown that a large-scale brain network encodes motor learning and transfer of learning from related past experiences (Heitger et al., 2012),(Gerraty et al., 2014). The prefrontal cortex (PFC) has been found to integrate the information necessary for action generation and perception (Raos and Savaki, 2017) relevant to error processing during FLS task performance. Specifically, FLS task performance is graded based on the speed and accuracy of the psychomotor skills (Ritter and Scott, 2007), where speed-accuracy tradeoff during skill training can lead to automaticity when there is a greater focus on speed despite residual error, i.e., an increased speed of action selection at the cost of cognitive flexibility (Poldrack et al., 2005; Toner et al., 2015) affecting error processing. Indeed, not everyone can achieve proficiency (Grantcharov and Funch-Jensen, 2009), and we postulate that successful skill acquisition needs cognitive flexibility (Poldrack et al., 2005; Toner et al., 2015) for error-based motor learning despite post error slowing of action selection (Seidler et al., 2013). Here, successful skill acquisition leads to an internal forward model (Wolpert et al., 1998) that can simulate the perceptual consequences of the planned and executed motor commands. An intact action-perception coupling has been shown to depend on the integrity of the cerebellum (Christensen et al., 2014) that underpins the internal model (Ebner, 2013) and error-based learning (Popa and Ebner, 2019). Error-based sensorimotor learning also involves other brain areas, including the parietal cortex, striatum and anterior cingulate cortex (Seidler et al., 2013). Then, the hierarchy of the cognitive control during skill learning shows a rostrocaudal axis in the frontal lobe (Badre and D’Esposito, 2009a), where a shift from posterior to anterior is postulated to mediate progressively abstract, higher-order control expected with skill learning. In this study, the dorsolateral and ventrolateral PFC showed activation in Fig. 7A,B during the FLS complex task that can be related to attention control, cognitive control, feature extraction, and formation of first-order relationships (Badre and D’Esposito, 2009b),(Badre, 2008),(Koechlin and Summerfield, 2007),(Christoff and Gabrieli, 2000). Specifically, the dorsolateral PFC of the dorsal stream is more involved in the visual guidance of action in novices (Fig. 7A) relevant in motor exploration (Sheth and Young, 2016). In contrast, the ventrolateral PFC of the ventral stream is more involved in the recognition and conscious perception (Milner, 2017) in experts (Fig. 7B) relevant in motor exploitation (Sheth and Young, 2016). Then, the supplementary motor area (SMA) and the premotor cortex are crucial for the coordination of bimanual movement (Tanji et al., 1988), where SMA is crucial for complex spatiotemporal sequencing of movements (Debaere et al., 2004),(Swinnen and Wenderoth, 2004) necessary in bimanual FLS complex task (Fig. 7B). Also, the cingulate and pre-supplementary motor areas are the generator sites of error-related negativity that is time-locked to an erroneous response (Seidler et al., 2013). Here, the medial frontal cortex is known to serve a central role in performance monitoring (Fu et al., 2022) that is crucial for cognitive flexibility. In this study, the dACC activity was captured by microstate 3 that was one of the most dominant (high GEV) microstates across all conditions. Future study needs to combine pupil dilation with EEG microstate analysis which may elucidate the relationship of the microstate 3 with the error-related pupil dilation during skill training vis-à-vis significance of errors for adaptive behavioral adjustments (Maier et al., 2019). Then, SMA is involved in planning complex motor finger tasks (PE et al., 1980) that is critical in error correction (Seidler et al., 2013). Here, EEG microstates, e.g., related to the canonical subjective interoceptive-autonomic processing (Britz et al., 2010), may be a marker of error specific autonomic arousal mechanisms that promote post-error adjustments (Maier et al., 2019) differentially in fast versus slow learners.
Brain-behavior monitoring of the error-related cortical activation and corrective action can allow appropriate error feedback for operant conditioning in future work that has been shown feasible in our prior application for stroke rehabilitation (Kumar et al., 2019). For example, novices may lack error perception (e.g., lack of medial frontal cortex activation on errors (Gehring and Fencsik, 2001)) that can disrupt their skill learning, which can be improved with non-invasive brain stimulation of the medial frontal cortex in conjunction with explicit error feedback in the medical simulator. Then, EEG topographies provide subject-specific correlates of motor control (Pirondini et al., 2017), where portable neuroimaging guided non-invasive brain stimulation may be feasible (Walia et al., 2021) to enforce beneficial scalp topographies to facilitate perception and action that together form a functional system. The two crucial attributes of the perception-action cycle are perceptual and executive memory (Fuster, 2004), and error sensitivity is postulated to depend on the memory of errors, i.e., the history of past consistent perceptual errors, e.g., error in depth prediction from a 2D view (Popa and Ebner, 2019) or executive errors, e.g., "incorrect needle insertion" (Albert et al., 2021). Then, early efferent error prediction can lead to fast adjustments in experts who know the action semantics, e.g., skilled typists execute errors with lighter keystrokes than novices. Published studies have shown that the pre-supplementary motor area (pre-SMA) and the inferior frontal gyrus are involved in stop-signal task performance (Seidler et al., 2013) necessary for immediate error-related adjustments. Also, published fNIRS studies showed the involvement of the inferior parietal cortex, PFC, occipital cortex, and the sensorimotor areas, including the premotor and primary motor cortex during skill training while the fMRI studies showed additional activation of deeper brain structures, including the basal ganglia and cerebellum (Roberts et al., 2006). The current study used portable fNIRS that has a limited spatial and depth sensitivity (Strangman et al., 2013) so could provide a partial view of the brain network.
The limitation of our study includes a low-density fNIRS and EEG sensor montage that limited the spatial resolution to capture the complete hierarchy of cognitive control in experts and novices. It is known from skill training studies that the hierarchy of cognitive control shows a rostrocaudal axis in the frontal lobe where a shift from posterior to anterior is postulated to mediate progressively abstract, higher-order control. Our multimodal imaging approach also limited the head cap space for each of the modalities due to separate optodes and electrodes in the sensor montage, where an integrated "co-located" optode + electrode (optrode) can be helpful (Keles et al., 2016) for high-density brain imaging in the future.