Functional near-infrared spectroscopy-based neurofeedback training induces changes in cortico-striatal functional connectivity: A proof of concept study

doi:10.21203/rs.3.rs-3145658/v1

Download PDF

Article

Functional near-infrared spectroscopy-based neurofeedback training induces changes in cortico-striatal functional connectivity: A proof of concept study

https://doi.org/10.21203/rs.3.rs-3145658/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 28 Aug, 2024

Read the published version in Scientific Reports →

You are reading this latest preprint version

Due to its central role in cognitive control, the dorso-lateral prefrontal cortex (dlPFC) has been the target of multiple brain modulation studies. In the context of the present pilot study, the dlPFC was the target of 8 repeated neurofeedback (NF) sessions with functional near infrared spectroscopy (fNIRS) to assess the brain response during NF and with functional and resting state magnetic resonance imaging (task-based fMRI and rsMRI) scanning. Fifteen healthy participants were recruited. Cognitive task fMRI and rsMRI were performed during the 1st and the 8th NF sessions. During NF, our data revealed an increased activity in the dlPFC as well as in brain regions involved in cognitive control and self-regulation learning (pFWE < 0.05). Changes in functional connectivity between the 1st and the 8th session revealed increased connectivity between the posterior cingulate cortex and the dlPFC, and between the posterior cingulate cortex and the dorsal striatum (pFWE < 0.05). Decreased left dlPFC-left insula connectivity was also observed. Behavioural results revealed a significant effect of hunger and motivation on the participant control feeling and a lower control feeling when participants did not identify an effective mental strategy, providing new insights on the effects of behavioural factors that may affect the NF learning.

Physical sciences/Physics/Techniques and instrumentation/Spectroscopy

Biological sciences/Neuroscience/Cognitive neuroscience

Biological sciences/Neuroscience/Computational neuroscience

Biological sciences/Neuroscience/Emotion

Biological sciences/Neuroscience/Feeding behaviour

Biological sciences/Neuroscience/Motivation

Biological sciences/Neuroscience/Neural circuit

Neurofeedback

dorsolateral prefrontal cortex

functional near infrared spectroscopy

functional connectivity

magnetic resonance imaging

eating behaviour

Due to its major role in executive functions ^[1], particularly in working memory ^[2], the dorsolateral prefrontal cortex (dlPFC) is a key region involved in the regulation of cognitive and behavioural functions, such as higher-order executive components of behavioural planning, including monitoring, manipulation, and integration of information ^[3]. Localized in the lateral prefrontal cortex, the dlPFC covers Brodman Area (BA) 9, 46 and 9/46 ^{[4, 5]}. The dlPFC is part of the cortico-striatal circuitry and of the prefrontal connectome ^{[6, 7]}. The cortico-striatal connections are highly involved in learning and executing optimal goal-directed behaviour, combining their emotional, motivational, cognitive and motor dimensions ^[6]. The dlPFC is thus functionally connected with near and distinct brain regions grouped in networks that underlie cognitive and behavioural processes. Additionally, the dlPFC is a key node of the Fronto Parietal Network (FPN), which plays an important role in cognitive control ^[8]. This network also includes dorsomedial PFC, supramarginal gyrus (SMG), posterior parietal cortex (PPC), the dorsal caudate and anterior thalamus, and is referred to as the “executive control network” (ECN) ^{[8, 9]}. This network underlies executive functioning and goal-oriented behaviours, interacting with the default mode network (DMN) through the salience network (SN) ^[10], where the posterior cingulate cortex (PCC) and the insula with the anterior cingulate cortex (ACC) are key brain regions of these circuits respectively. Thus, there is a wide interplay between dlPFC and other distinct regions from different functional networks. Accordingly, disruption in dlPFC’s connectivity has been shown to play a role in a wide range of cognitive and behavioural traits and disorders, including those in relation with eating ^{[11, 12]}.

In the scope of food intake regulation, the dlPFC has been described as a key actor by exerting an inhibitory control on impulsive and unhealthy food choices (for review, Val-Laillet et al., 2015). Overeating is related to impulsivity and deficits in inhibitory control ^{[14, 15]}, even in individuals with a healthy body mass index (BMI) without eating disorders ^[16]. In addition, as the dlPFC also plays a key role in the cognitive control of emotion regulation ^[17], it can be involved in emotional overeating ^[18]. Beyond the dlPFC, some others brain areas are of interest in the cognitive control of (emotional) eating behaviour, namely the dorsal striatum (caudate and putamen) which is involved in feedback processing and in the mediation of habitual behaviour when exposed to food cues ^[19]. Brain areas involved in reward processing and food valuation systems (ie. the ventromedial prefrontal cortex, vmPFC) (Hollmann et al., 2013; Rolls, 2008; Shin et al., 2009), interoceptive awareness (ie. the insula) (Craig, 2009; Shin et al., 2009), and PCC for its role in memory and learning ^[19] are also elicited. All of these brain regions display functional connectivity (FC) with the dlPFC ^[23]. Overall, self-control of food choices depends on a prefrontal network centred on the dlPFC interacting with reward-related brain regions ^{[24, 25]}. While decreased BOLD responses in the dlPFC/middle frontal gyrus or reduced fronto-striatal connectivity have been highlighted in individuals suffering from craving and/or obesity ^{[16, 26, 27]}, higher dlPFC activity is associated with higher self-control in food-related decision-making (Chen et al., 2018). Consequently, our hypothesis is that up-regulating the dlPFC might be of interest for preventing food craving and overeating.

To this aim, neurofeedback (NF) training can be implemented to help a subject learn to self-regulate the activity of a brain region or the FC between different brain regions. The measured brain activity is instantaneously retransmitted to the subject through a brain computer interface (BCI) allowing him/her to apply a volitional control over it ^{[29, 30]}. Thus, NF is a kind of biofeedback, which provides a conscious information about an unconscious physiological phenomenon ^[31]. Based on learning skills ^[30–33], some psychological and behavioural factors may influence NF learning success, such as the mental strategies employed to gain control over the feedback ^[34–36], well-being or motivation ^{[31, 37]}. Since it is well known in NF literature that 15–30% of individuals can be classified as “non-responders”, referring to the inability to regulate their own brain activity, controlling for such factors is recommended ^[36]. The dlPFC has been the target of various NF protocol designs in a wide range of population with fMRI ^[38–40] or with functional Near-InfraRed Spectroscopy (fNIRS) ^[41–47]. Moreover, some fMRI-based NF studies directly targeted the FC between the dlPFC and other brain regions such as the insula ^[48], the anterior cingulate cortex (ACC) (Morgenroth, 2020), or within the prefrontal-striatal network ^[49]. In the context of eating behaviour, one study targeted the dlPFC in obese individuals ^[40] and another study targeted the FC between the dlPFC and vmPFC in overweight and obese individuals ^[50], in order to improve self-control toward healthier food choices. Of interest, two dlPFC-targeted NF studies reported changes in the cortico-striatal regions ^[40] and within the emotion regulation network (ERN) and FPN ^[47], suggesting that targeting the dlPFC may be efficient to provide changes in brain networks.

In this pilot study, we hypothesized that a NF training targeting the dlPFC would increase cortico-striatal FC in healthy individuals. In order to characterize the NF brain effects on the whole-brain, we implemented a bimodal acquisition with MRI during the 1st and last (8th ) fNIRS-based NF sessions. Psychological factors that could influence NF learning outcomes (e.g. motivation, well-being/mood, perception of self-control) as well as the mental strategies employed by the participants were also assessed. We investigated whether this training was associated with subjective control feeling from the participants and its relation with other collected behavioural measures.

Behavioural Results

The descriptive statistics of the behavioural measures are reported in the Supplementary Table 1. Time effect through GLM analyses did not reveal any changes across time regarding the motivation (assumed sphericity, D=0.712; ddl=7; p=0.662), the intensity of the effort provided (Greenhouse-Geisser correction, D=0.403; ddl=2.73; p=0.733), the well-being (assumed sphericity, D=1.018; ddl=7; p=0.424), or the gauge control feeling (assumed sphericity, D=1.831, ddl=7; p=0.091). When exploring the variables that may have affected the gauge control feeling, i.e. hunger, thirst, well-being motivation and effort intensity regarding NF, univariate GEE analysis revealed an effect of hunger (B=-0.177; p=0.034) and an effect of motivation (B=0.277; p=0.028). In a second step, a multivariate GEE analysis was performed by integrating the motivation and hunger state (motivation, B=0.231; p=0.048; hunger, B=-0.071; p=0.03). The results of the GEE analyses are summarized in Table 1. With regard to the strategies used, the overall distribution in percentage is represented in Fig 1.a. The distribution of the strategies used per session and per participant is represented in Fig 1.b. The gauge control feeling was significantly different between strategies (One-way ANOVA, F=10.209; ddl=3; p<0.001, Fig 2C). Post-hoc analyses showed that the mean gauge control feeling associated with “No strategy” was significantly lower than that associated with the other strategies. No significant difference was observed between the other strategies (see supplementary Table 2).

fNIRS-based activation

Overall, 74% of the channels of interest reached the SCI correction threshold. The proportion of individuals having at least one activated channel covering the dlPFC for each NF session are expressed in percentage in Fig 2.a. We observed that 12 out of the 15 participants had at least one channel of interest activated during at least 5 NF sessions. At the second level merging the 1^st and the 8^th sessions, we observed that 2 out of the 4 channels covering the upper bilateral dlPFC (S1-D2 and S7-D5 channels; pbonf<0.05) and two channels covering the middle frontal area (S4-D5 and S4-D2) were significantly activated (Fig a.b). Beta-values and z-score are summarized in Table 2. Activation maps of the second level analysis for each NF sessions are also represented in the Supplementary Fig S1. Beta-values and z-score are summarized in Supplementary Table 3.

fMRI brain responses during the 1^st and 8^th Neurofeedback sessions

The activation map revealed activated brain clusters in cortical areas, including the dlPFC, and subcortical areas. According to our a priori hypothesis, SVC analysis performed on bilateral dlPFC revealed a bilateral significant increased activity (pFWE<0.05). The SVC analysis also revealed bilateral activation in IFG, PCG, insula, SMA and SMG and. Significant activations were also found in the cingulate cortex (e.g. ACC and middle cingulate cortex, MCC), nuclei of the thalamus (ventro-lateral and ventral-anterior) and some regions of the basal ganglia with dorsal striatum (caudate, putamen) and pallidum, which survived correction for multiple comparison (pFWE<0.05) and are represented on the statistical maps in Fig 3. The statistical results are summarised in Table 3.

Resting-State Functional Magnetic Resonance Imaging-Based Functional Connectivity

The effect of NF training on the FC was measured between the eighth versus the first NF sessions. Statistically significant functional connectivity differences were found between the NF sessions after false discovery rate (FDR) correction and are summarized in Table 4. Overall, the ROI-to-ROI FC connectome ring on the positive side (Fig 4.a) revealed significant increased FC in 6 connections among the PCC, bilateral caudate, putamen and dlPFC. The ROI-to-ROI FC connectome ring on the negative side (Fig 4.b) revealed a significant decreased FC between the left insula and the left dlPFC within a cluster.

This proof-of-concept study consists in an overall characterization of an 8-session NF training over a month for further application in the context of emotional overeating. In this study, we demonstrated the feasibility of repeated fNIRS-based NF sessions targeting the dlPFC in 15 normal-weight healthy individuals. To our knowledge, it is the first dlPFC-targeted NF interventional study implementing an fNIRS and fMRI bimodal acquisition. Thanks to this approach, we demonstrated that the participants were able to engage specifically the activity of their dlPFC, among other cortical and subcortical brain regions. In addition, we demonstrated that 8 NF sessions over a month could induce changes in the corticostriatal functional connectivity. We also investigated the behavioural factors and individual mental strategies that may affect the NF learning, which will be discussed below.

Regarding the subjective gauge control feeling, we found a positive effect of the participants’ motivation to perform the NF task, suggesting that increased motivation was associated with higher control feeling. We also observed a negative effect of hunger status prior to the NF sessions, suggesting that higher hunger self-assessment was related to lower self-reported control feeling. Computed together in the multivariate analysis, the effect of motivation and hunger remain significant but to a lesser extent for the motivation, suggesting a possible mediating effect of the hunger state between control feeling and motivation. This result is in line with the literature on the impact of behavioural and psychological factors on NF learning, especially motivation ^{[31, 37]}. Indeed, the motivation level has been suggested as a moderate predictor of learning success by increasing the focus on the task and may be linked to the attentional resources ^[37]. We suggest that hunger may have focused the participants’ attention towards their visceral sensations during NF learning, rather than towards the NF task, consequently acting as an attentional distractor. Overall, hunger may have affected attentional resources allocated to the gauge by decreasing motivation. Therefore, we further suggest avoiding fasted state before every NF session. Note that the self-reported well-being of the individuals as well as the intensity of the effort provided remained stable over time, suggesting that the NF interventional study was not detrimental to the psychological well-being of the participant and also that they did not disengage from the study. Regarding the relationship between control feeling and the mental strategies employed, no particular strategy was more effective than the others. However, our results suggest that having identified one’s own successful strategy is better than having none, as the mean control feeling with “no strategy” category was lower as compared with the other categories. This result comforts the idea not to provide any specific instruction to the participants to get a volitional control over their dlPFC activity and to let them find by themselves their most efficient strategy. This is in line with the NF learning theories ^{[30, 33, 51–53]}, where NF relies on implicit learning, principally based on operant conditioning, through series of trials and errors. Note that our "no strategy" category included NF sessions in which an individual did not try any strategy and also those who tried many but did not find any effective strategy. As a result, the “no strategy” category included individuals who tried several strategies, suggesting that a significant cognitive load was recruited for some participants. It has been suggested that using lots of strategies may reflect a lack of automatization and may lead to an overload of cognitive resources ^{[34, 35]}. This is in line with the conclusion of Kober et al. (2013) and may explain why in our study having no efficient strategy was associated with lower control feeling reflecting overloaded cognitive resources. It is worth noting that the strategies were compared to subjective ratings and not to an objective marker of NF success, such as the variation in HbO amplitudes with fNIRS ^[29]. Therefore, the conclusions of our study should be taken cautiously.

For the fNIRS results, we found that most of the individuals were able to engage their dlPFC although discrepancies exist between individuals. Some of them (4, 26.6%) showed a significantly increased dlPFC activity for each NF session, some others (3, 20%) showed a significant increase in their dlPFC activity in up to 4 out of the 8 NF sessions. In the second-level analysis, in line with our fMRI results, we found an increased activity of the dlPFC with an activation of the upper central frontal region. Since a quarter of the channels targeting the dlPFC did not reach SCI threshold, we cannot exclude the fact that the dlPFC activity retransmitted to the participant via the metaphor may have been affected by noise. Nevertheless, in a fMRI study, Ninaus et al. (2013) investigated the neural processes involved in cognitive control during a sham NF (where the NF consisted in a random signal but individuals thought it was related to their actual brain activity). The authors found that the intention to control the retransmitted signal was enough to engage a broad frontoparietal and cingulo-opercular networks involved in cognitive control (^[54]) even if the feedback did not reflect real brain activity. Based on this conclusion, since our participants all knew that the feedback reflected their dlPFC activity (as there was no sham group), noisy channels is not likely to have hindered their NF learning processes since they were engaged in the training as reflected by the absence of change in the effort intensity provided. Here we would like to emphasize the importance of quality control for further fNIRS-based NF studies, as in a recent review, among the 100 most cited fNIRS-based BCI studies, approximatively 40% reported applying SNR pruning of their data ^[55].

In line with our fNIRS results, fMRI analysis also revealed increased activity in the dlPFC. Our results are in agreement with three fMRI-based NF studies where individuals successfully upregulated their left dlPFC activity ^[38–40], although these studies had different experimental paradigms and data analysis approaches. In addition, we also found similarities in brain activation with Kohl et al. (2019), who reported a large significant cluster in the bilateral anterior insula extending to bilateral IFG, thalamus and dorsal striatum and the right dlPFC when comparing NF task vs rest. Indeed, beyond the dlPFC, we also found that different cortical and subcortical areas were recruited, of which many display structural and/or functional connectivity with the dlPFC ^{[6, 7, 23]}. We notably found an increased BOLD activity in regions of the basal ganglia, including the pallidum and dorsal striatum (bilateral caudate and putamen). The dorsal striatum is involved in learning reinforcement based on feedback/reward processing, especially in different aspects of motivational and learning processes that support goal-directed actions, as well as in decision-making and in the control of habitual actions ^[56–58]. In addition, it has been shown that putamen volume predicts NF learning success ^[59] and that stronger activation of the striatum is associated with successful NF self-regulation ^[60]. Taken together, our results are in line with the literature where the striatum plays a key role in NF learning ^[30]. We also found an increased activity in bilateral insular cortex, which is important for the meta-representation of interoceptive signals and awareness (Craig, 2009), and might reflect interoceptive attention allocated to the internal signals during the NF sessions.

We also found increased activity in subregions of the cingulate cortex, including the ACC, which plays a role in attentional and emotional processes and in representation of reward value and punishments, and the MCC, which is mainly involved in response selection and feedback-guided decision making ^[61]. Both ACC and MCC are involved in error detection made in many tasks (Rolls, 2019). Since the NF learning is mainly based on series of trials-and-errors, we hypothesized that the cingulate cortex exerts a major role on the attention allocated to successes and failures to control the gauge and in subsequent decision-making processes to gain voluntary control over it. We also found increased activity in areas involved in motor functions, such as the PCG, SMA, putamen and middle frontal gyrus (BA 9/46, 6, 8, and 32). Activity of these regions have been highlighted during mental imagery of motor movements in a recent meta-analysis, where the increased dlPFC activity was thought to be related with the cognitive demand of working memory ^[63]. Since many participants used mental strategies during the NF sessions, notably related with mental visualization, these regions might be activated in response to these type of mental strategies. We also found a bilateral activation in the IFG. This region, in particular its left part, is related to working memory, language and phonological processing ^[64–66], including inner speech in self-reflective processes ^[67]. As a mental strategy, some participants sang in their heads, recalled lyrics from music or poem-like texts and went over mental to-do lists. Since we found a cluster in left BA 44 and 45, we hypothesized that IFG recruitment during the NF may be related to these functions ^[64]. In addition, we also found increased activity in the thalamus, which has been proposed to be involved in the “control network”, comprising lateral occipital cortex, dlPFC, and posterior parietal cortices (PPC), where dlPFC and PPC are connected with the thalamus in order to regulate cortical arousal during NF ^[30]. Although the dlPFC was our target ROI for the NF paradigm, this brain region seems to play a key role in NF learning. Indeed, its activation has been highlighted in a meta-analysis in which the authors investigated the neural markers underlying brain regulation during NF ^[68]. This meta-analysis also revealed ventrolateral areas of the prefrontal cortex, basal ganglia, insula, ACC, thalamus and visual associative areas ^[68]. Taken together, beyond successful dlPFC upregulation, our results suggest that our designed NF paradigm can successfully recruit brain regions involved in the cognitive processes of learning brain self-regulation (encompassing working memory and attentional processes), mainly from the cortical-basal ganglia loops responsible for procedural learning ^[53].

The rsMRI analysis revealed significant changes in the FC of key nodes of the CEN (dlPFC), the DMN (PCC) and in the SN (insula). Indeed, we found increased FC between the PCC and bilateral dlPFC, and between bilateral putamen and caudate. The PCC has been suggested to play a key role in learning for environmental change detection and subsequent alterations in behavioural policy ^[69]. Indeed, in a model of change detection and policy selection sensory feedback from reward outcome, it has been proposed that the PCC interacts with the dlPFC and with basal ganglia for encoding individual event-related outcomes necessary for altering behaviour ^[69]. In addition, the PCC has been shown to play a role in basic cognitive processes typically suppressed during performance of well-learned tasks ^[69]. This result is in line with growing evidence about the role of the PCC in cognition and attention, especially supporting internally directed cognition ^[70]. This increased dlPFC-PCC connectivity and PCC-dorsal striatum might support increased internally oriented attention, cognitive control and learning. In addition, in a NF study targeting the right dlPFC, an increased functional connectivity with FPN was found, and also within the emotion regulation network at the end of training, supporting evidence for enhanced interaction between cognitive control and emotion processing ^[47]. In line with our results, this emphasizes the benefits of dlPFC-targeted NF training to increase the functional connectivity in networks supporting cognitive control. We also found significant decreased functional connectivity between the left insula and left dlPFC. In a study using repetitive transcranial magnetic stimulation (rTMS) of the dlPFC in depressive patients, a decreased functional connectivity between the left dlPFC and left insula was described at the end of the intervention in the rTMS groups as compared to the control group, which correlated with symptoms improvement (Struckmann et al. 2022). In addition, Kohl et al. (2019) found a stronger negative FC between the dlPFC and right insula after neurofeedback training targeting the left dlPFC. Stronger coupling between these regions correlated with a greater decrease in the rated palatability of high-calorie foods items ^[40]. However, it is worth noting that, unlike in our study, the analysis was performed using a weighted GLM to take into account the HRF, as the FC was assessed during the NF to define specific weights within conditions. In our study, the decreased connectivity between the 1st and 8th sessions could reflect decreased need for awareness over time. It would be of interest to investigate the link between this FC and behavioural assessments in emotional overeaters in a future clinical NF study, as the insula also plays a role in emotional processing ^[71–73] and reward processing related to eating behaviour ^[19].

There are several limitations to this study. We included only 15 participants. Thus, our results should be considered as a proof-of-concept for further studies in clinical populations and are not generalizable outside our restrained population. Nevertheless, FC has been assessed in a NF study with the same sample size, yielding significant and valuable results ^[74]. Regarding our NF paradigm, some limitations should also be taken into consideration. We did not perform an individualized localizer run for dlPFC localization but rather chose the same fNIRS channels covering the whole dlPFC brain region. Thus, we cannot exclude anatomical differences between participants, leading to small differences in dlPFC coverage. In addition, regarding the learning success of the NF training, we investigated the subjective gauge control feeling and did not apply a transfer run to actually evaluate whether the individuals were able to perform the NF task without real-time feedback, reflecting a successful learning of brain self-regulation. Therefore, we can only conclude that our NF paradigm did involve the recruitment of brain regions related to cognitive control and working memory, but we cannot confirm a learning success. Moreover, since this study was a proof-of-concept study, we did not include a control group (e.g. sham NF), which is strongly recommended for application in clinical populations with expected behavioural outcomes ^{[32, 36, 75, 76]}. Finally, this study should be considered in the light of its strengths. We followed the recommendations for best practice in the use of the fNIRS for real-time NF ^[77] and controlled for good SNR, limiting false-positive misinterpretation. In addition, we followed most of the recommendations from the CRED-CHECKLIST ^[36], by controlling for psychosocial measures, reporting mental strategies, taking into account online data processing and artefact correction, providing feedback specifications and reporting brain outcomes measures. Due to these quality checks, we believe that this study will contribute to increasing the general knowledge about the brain effects of a repeated NF training protocol targeting the dlPFC. In addition, no major harm nor unintended effects were reported in the present study, confirming the safety of this non-invasive approach. To conclude, in light of the results of this study, we would like to highlight the potential of dlPFC-targeted NF training to improve cortico-striatal connectivity and emphasise on its potential to improve cognitive control in a target population, for example in the context of eating behaviour.

Ethics statement

This pilot study was sponsored by Rennes University Hospital (N° ID-RCB: 2021-A02314-37) and approved by the Institutional Review Board (ClinicalTrials.gov Identifier: NCT05200182, https://clinicaltrials.gov/ct2/show/NCT05200182; First trial registration: 20/01/2022). The study was conducted in accordance with the Declaration of Helsinki and with the French and European General Data Protection Regulation 2016/679. Free written consent from all the participants was collected before inclusion in the study. Patient identification was removed and all data were pseudonymised. No significant changes were made to the methods after the start of the pilot test, nor were any changes made to the pilot test assessments or measurements.

Participants

Fifteen normal-weight adults (9 women, 6 men, mean age = 28.7 years ± 7.2, mean body mass index BMI = 21.5 ± 1.3; mean ± SD) were recruited to participate in the study following a flyer distribution campaign within the personal network of the institute members. No other demographic data was collected. This pilot study consisted in an 8-session fNIRS-based NF training protocol. The recruited participants were naive about the NF training, had a normal or corrected-to-normal vision and were right-handed (to avoid the inter-individual variability related to brain laterality). Each participant had a sufficient level of French language to understand the implications of the project and the instructions provided by the experimenters during the study. Each participant reported orally no history of neurological or neuropsychiatric disorders. They perceived 160€ as a financial compensation for their participation in the study.

Study procedure

The participants performed 8 fNIRS-based NF sessions across 4 weeks, with 2 sessions per week. Each NF session was spaced with at least one day without training. The NF task was identical from session to session but the first and last sessions were performed during MRI in a lying position, while from the 2nd to the 7th session, the participants were installed in a sitting position in a dedicated room to perform the NF task with fNIRS only (see overall design study in Fig. 5. a). Before and after each NF session, the participants had to fill in a behavioural questionnaire (see below). The study began in early April 2022 and ended in August 2022 at the last NF session of the 15th participant as 15 inclusions were expected for this pilot study. No follow-up was foreseen. There were no lost or excluded participants. First and eighth NF sessions. Participants were asked to come to the appointment with a minimum fasting period of 2 hours. The NIRS cap was installed on the participant’s head. The standardised instructions related to the NF task and the steps of the MRI acquisitions were provided orally. For the first NF session, the participants underwent an rsMRI acquisition phase before performing the NF. The NF was displayed on a LCD computer screen (NordicNeuroLab, Bergen, Norway) placed at the back of the bore and visible to the participant via a mirror mounted on the head coil. The 8th session was identical to the first one, except that the sequence started with the NF and ended with rsMRI to ensure that the last rsMRI acquisition was performed after all of the 8 NF sessions.

From the 2nd to 7th neurofeedback session. The participant sat in front of an LCD computer screen on which the NF task was displayed. Then, the NIRS cap was placed on his/her head. The experimenter explained the instructions before sitting behind the participant’s screen, then being no longer in the participant's field of view.

Self-report questionnaires

Before and after each NF session, the participants were asked to rate their hunger and thirst states, their current well-being and motivation to perform the NF with the use of Likert scales ranging from 1 to 10. At the end of the session, they also had to i) estimate the intensity of the effort made, ii) give their gauge control feeling and iii) describe the mental strategies they employed to control the gauge and identify the most effective strategy if any. They were encouraged to focus on the most effective strategy identified during the different NF sessions. All of these behavioural assessments were collected through self-reported written questionnaires. The mental strategies were classified from the self-declared strategy report by 3 blinded collaborators into 6 categories: Concentration (the participant employed a strategy that requires precise concentration/attention to raise the gauge), visualization (the participant visualized/imagined something/scene/memory to raise the gauge), cheering (the participant used a positive reinforcement/self-encouragement method to raise the gauge), relaxation (the participant used a relaxation/meditative method in order to raise the gauge), motor (the participant moved or contracted part of his/her body to raise the gauge) and no strategy (the participant was not able to identify any effective strategy or did not attempt to raise the gauge in any way).

Neurofeedback task

The NF paradigm included the use of a visual metaphor consisting in a vertical visual gauge, which represented the oxy-hemoglobin (HbO) concentration levels of the participants’ bilateral dlPFC. From the bottom to the top of the gauge, the colour was fading from yellow to red. One NF session lasted 17 minutes. After a baseline period of 2 minutes, NF and rest periods were shown alternatively, each lasting 30 seconds and with 15 iterations (Fig. 5.b). The participants were told that the dlPFC is a brain region involved in cognitive control and in goal-directed behaviours. They were then asked to try the mental strategy of their choice related to cognitive control in order to raise the gauge. They were advised to keep a steady breathing rate and to avoid any body or head movement. We chose to provide only implicit instructions not to prevent subjects from finding and testing their own strategies through self-learning processes ^[33]. During rest periods, participants were asked to let their mind wander to contrast with NF runs while limiting mental fatigue. To help them relax, a concentric blue circle alternatively increasing and decreasing in diameter (0.1 Hz, fixed duration, not related to measured signals) appeared on their computer screen and they were asked to keep a constant breathing rhythm.

Functional near-infrared spectroscopy (fNIRS) system

The MRI compatible NIRSCOUT system (NIRScout XP, NIRX, Berlin, Germany) was used with laser sources for all NF sessions under the MRI. The LED sources were used for the 6 ‘fNIRS only’ NF sessions for 8 participants with the NIRSCOUT system and the NIRSPORT system with LED sources was used the 7 others (NIRSport 1, NIRx, Berlin, Germany). The set-up consisted of a pairing of 8 sources and 7 APD detectors, where 16 channels covered the frontal cortex and a separate channel was located on the left motor area. Four channels targeting the bilateral dlPFC were selected as our channels of interest (Fig. 5.c).

Online NIRS analysis for real-time neurofeedback. The analysis was inspired by (Godet, Serrand et al. submitted in Plos One, minor revisions). Wavelengths were at λ_1|2 = 760|850 nm for the LED system and λ_1|2 = 785|830 nm for the LASER system and the sampling rate at 7.81 Hz. For signal acquisition, we used the NIRSTAR software (v15.2, NIRx, Berlin, Germany) and Lab Streaming Layer LSL (San Diego, USA, https://zenodo.org/record/6387090), a homemade software in Python to compute the NF feature and Psychopy ^[78] to display the NF task. In particular, following the Beer-Lambert Law ^[79], the raw signal from NIRSTAR was converted into [HbO] and [HbR] concentrations. A low-pass filter (Butterworth, order 3, f_c=0.09 Hz) and detrending of the four channels of interest were applied to the data. The mean value of these four channels was computed and subtracted by the mean signal value of the 5 seconds preceding the NF task to account for signal baseline.

Offline NIRS analysis. Block averages and activation maps were obtained with MNE-NIRS based on the Python package MNE-Python ^[80]. The raw data were converted into optical density, then bandpass-filtered (BP filter: 0.01Hz to 0.09Hz) before being converted into [HbO] and [HbR] concentrations using the modified Beer-Lambert law. A generalized linear model (GLM) based on the Nilearn Python package ^[81] was used to model the NF task-related hemodynamic response at the individual level. NF task and rest onsets were added to the GLM model with orthogonalized age entered and the NF task onsets were used for activation analysis. The default canonical hemodynamic response function (HRF) from SPM and implemented in Nilearn was used. To improve the signal-to-noise ratio, we also applied a scalp-coupling index (SCI) threshold at 0.8 in order to exclude noisy data ^[82]. Channels that did not reach this threshold were removed from the analysis to avoid false-positive misinterpretation. For each participant and for each brain region of interest (i.e. 4 channels covering the bilateral dlPFC, 16 channels covering the other frontal regions and 1 channel for the left motor region), we computed the number of activated channels as defined by a positive estimator beta-value reaching a Bonferroni corrected statistical threshold of p_Bonferroni<0.05. Given the limited sample size of this proof-of-concept study and to reach a sufficient statistical power for the fMRI analysis, we performed a second level analysis by merging the data from the 1st and the 8th NF sessions. For each channel, the β and t-values were computed.

MRI data acquisition

Functional and structural MRI images were obtained with a 3T MRI scanner (Magnetom Siemens Prisma, VE11C) equipped with a 64-channel head coil. For the 3D T1-weighted anatomical scan, the following parameters were used: Repetition time (TR) = 1900 ms, inversion time = 900 ms, echo time (TE) = 2.26 ms, flip angle = 9°, field of view = 256 x 256 mm2, 176 slabs leading to a voxel size of 1 x 1 x 1 mm³. fMRI and rsMRI images were acquired with a multiband echo-planar imaging sequence developed by the Centre for Magnetic Resonance Research (CMRR) at the University of Minnesota using the following parameters: TR = 1224 ms, TE = 30 ms, adjusted flip angle = 65°, matrix size = 84 x 84, multiband factor 3, no iPAT (integrated Parallel Acquisition Techniques), and 54 slices with a thickness of 2.5 mm, resulting in a voxel of 2.5 x 2.5 x 2.5 mm3.

For the fMRI sequences acquired during the 1st and 8th NF sessions, the pre-processing steps and analyses were performed with the Statistical Parametric Toolbox (SPM) 12 (Wellcome Department of Imaging Neuroscience, London, United Kingdom) in MATLAB (Mathworks, Inc., Sherborn, MA, United States). Pre-processing steps consisted first in normalization and smoothing of structural data (voxel size: 1x1x1mm). Then, the realignment of functional images to the mean to correct for motion correction artefacts from the structural 3DT1 images and normalization to the Montreal Neurological Institute (MNI) space were performed, resulting in a final voxel size of 2.5x2.5x2.5mm. Functional scans underwent slice-timing correction. Finally, spatial smoothing with a 6-mm full width at half-maximum kernel (FWHM) was performed. Head motion did not exceed 2mm in translation, and the sample size remained at 15 participants. At the individual level analysis, a GLM was performed on brain-masked data (Intracranial Volume mask, ICV), using the canonical HRF without time and dispersion derivatives. Realignment parameters were entered as covariates. Statistical maps were generated by convoluting the time series with the canonical HRF, modelling explicitly the active and implicitly the resting periods of the NF task. For a statistical power purpose, images acquired during the first and eighth NF sessions were summed to identify brain regions that were more responsive to NF for each individual. Then, for the analysis of all subjects, a two-sample t-test was performed with orthogonalized age entered as a covariate without the independence parameter to account for repeated measurements for the same individuals for both NF sessions. Small volume correction (SVC) analyses were performed to adjust for the number of voxels within the dlPFC using coordinates from ^[83] to address our a priori hypothesis and to refine ROI from the clusters identified with the xjView toolbox (http://www.alivelearn.net/xjview) and bSPMview (https://www.bobspunt.com/software/bspmview/). Also with SVC analysis, only brain activations that survived FWE (p < 0.05) correction for multiple comparisons at the cluster level and with a puncorr < 0.001 at the voxel level within the ROI identified in the clusters are reported. MNI coordinates for the voxels of maximum activation were identified using the AAL3v.1 atlas to confirm the anatomical location of each locus.

For the rs-fMRI acquisitions, the sequence lasted 20 minutes during which the participants had to let their mind wander and to keep their eyes opened. The acquisition procedure was similar as previously described in ^[84]. Pre-processing steps and data analyses were performed with the CONN Toolbox (https://web.conn-toolbox.org/) version 21.a on MATLAB 2018a environment (The MathWorks, Inc., Natick, MA, United States). The pre-processing steps included normalization to MNI atlas space, spatial smoothing (8-mm FWHM), motion artefact detection and denoising. At the second-level, through a ROI-to-ROI Connectivity approach (RRC, ^[85], ROI-to-ROI analysis was conducted to examine FC changes between the 1st and the 8th NF session (V8 > V1 contrast). Based on our a priori hypothesis (i.e. increased cortico-striatal connectivity after the 8-session NF training), ROIs in this matrix were selected based on ROI-to-ROI functional similarity metrics described in Jung and Ralph (2022) with key brain regions involved in eating behaviour and from networks altered in disordered eating behaviours ^[86]. Thus, we predefined 10 ROIs in the CONN toolbox: Bilateral dlPFC ^[83] as our NF target and key node of the FPN, bilateral caudate and bilateral putamen for the dorsal striatum, anterior cingulate cortex (ACC) and bilateral insular cortex as key nodes of the SN and the PCC as a key node of the DMN. Since the spatial pairwise clustering statistics (SPC) is suited to an effect between an isolated pair of regions, whereas network based statistics (NBS) is better suited to an effect spanning multiple interconnected regions ^[87], we selected statistically significant ROI-to-ROI connections based on SPC statistics ^[88] with default settings in CONN. We employed the hierarchical clustering method (Conntoolbox default, code “0” for functional hierarchical clustering criteria). The T-statistics of the entire ROI-to-ROI matrix were estimated using a general linear model, where age was defined as the confounding covariate. Both possible one-tailed alternative hypotheses were tested separately (positive and negative sides). The statistical parametric map was thresholded using a significance level of p < 0.01 (uncorrected) and cluster-level FDR-corrected p < 0.05 was applied.

Statistics

Statistical analyses were conducted with the SPSS software (SPSS v20, Chicago, Illinois, United States). Questionnaire’s statistical analyses included descriptive statistics and the investigation of the effect of time on the four following variables: Gauge control feeling, motivation, well-being before the NF session and effort intensity. A repeated Generalized Linear Model (GLM) was conducted on these variables with age integrated as a covariate into the model. When the assumed Mauchly’s sphericity was violated, Greenhouse-Geisser correction was made. We also examined the factors (behavioural assessments and population’s characteristics) that may have influenced the variable related to the gauge control feeling. Since our study consisted in repeated scale assessments, we employed a generalized estimation equation technique with a first-order autoregressive correlation structure, which extends the generalized linear model to include analysis of repeated measurements or other correlated observations ^{[89, 90]}. We used a two-step approach, with a univariate analysis in order to estimate the strength of the association between the gauge control feeling and each covariate. All variables that demonstrated an association with p < 0.05 in the univariate analysis were considered as candidate variables in a multivariate model. Furthermore, in order to investigate the relation between the strategies employed and the gauge control feeling, a one-way ANOVA analysis was performed with post-hoc LSD test on strategies reported during the 120 NF session to investigate the differences between each type of strategy. The strategies related to motor (n = 1) or relaxation (n = 3) categories were excluded due to small number. For each statistical test, a p-value < 0.05 was considered significant and p-values in the range of [0.05–0.099] considered as a tendency.

Acknowledgments

We thank Aymery Constant for his advices about the GEE analysis. Data acquisition was performed at the Neurinfo imaging research facility from the University of Rennes, the University Hospital of Rennes, Inria, CNRS and the Rennes Cancer Center. Neurinfo is also supported by the Brittany Council, Rennes Metropole, GIS IBISA, CNRS INS2I and FEDER.

Author contributions

RM, NC, EB and DVL designed the study. RM enrolled participants. AG and BL performed the experimental acquisitions. AG, YS and NC contributed data analysis. AG and YS wrote the manuscript. All authors reviewed the manuscript.

Data availability statement

Raw data statistics are available at https://doi.org/10.57745/O0KPNY.

Funding statement

The present research was funded by the University of Rennes 1, Fondation de l’Avenir, the Benjamin Delessert Institute, and INRAE. A. Godet received a PhD grant from the University of Rennes 1. The fNIRS device used in this study was funded by CNRS INS2I and FEDER.

Conflict of interest disclosure

Authors declare no conflict of interest or competing interest. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Ethics approval statement

This pilot study was sponsored by Rennes University Hospital (N° ID-RCB: 2021-A02314-37) and was approved by the Institutional Review Board (ClinicalTrials.gov Identifier: NCT05200182, https://clinicaltrials.gov/ct2/show/NCT05200182). Informed consent was obtained from all participants.

Friedman, N. P. & Robbins, T. W. The role of prefrontal cortex in cognitive control and executive function. Neuropsychopharmacology 47, 72–89 (2022).
Diamond, A. Executive functions. Annual Review of Psychology vol. 64 135–168 at https://doi.org/10.1146/annurev-psych-113011-143750 (2013).
Tanji, J. & Hoshi, E. Role of the lateral prefrontal cortex in executive behavioral control. Physiological Reviews vol. 88 37–57 at https://doi.org/10.1152/physrev.00014.2007 (2008).
Rajkowska, G. & Goldman-Rakic, P. S. Cytoarchitectonic definition of prefrontal areas in the normal human cortex: II. Variability in locations of areas 9 and 46 and relationship to the Talairach coordinate system. Cereb. Cortex 5, 323–337 (1995).
Petrides, M. Lateral prefrontal cortex: Architectonic and functional organization. Philos. Trans. R. Soc. B Biol. Sci. 360, 781–795 (2005).
Haber, S. N. Corticostriatal circuitry. Dialogues Clin. Neurosci. 18, 7–21 (2016).
Haber, S. N., Liu, H., Seidlitz, J. & Bullmore, E. Prefrontal connectomics: from anatomy to human imaging. Neuropsychopharmacology vol. 47 20–40 at https://doi.org/10.1038/s41386-021-01156-6 (2022).
Menon, V. & D’Esposito, M. The role of PFC networks in cognitive control and executive function. Neuropsychopharmacology vol. 47 90–103 at https://doi.org/10.1038/s41386-021-01152-w (2022).
Seeley, W. W. et al. Dissociable intrinsic connectivity networks for salience processing and executive control. J. Neurosci. 27, 2349–2356 (2007).
Menon, V. Large-scale brain networks and psychopathology: A unifying triple network model. Trends in Cognitive Sciences vol. 15 483–506 at https://doi.org/10.1016/j.tics.2011.08.003 (2011).
Chen, S., Dong, D., Jackson, T., Su, Y. & Chen, H. Altered frontal inter-hemispheric resting state functional connectivity is associated with bulimic symptoms among restrained eaters. Neuropsychologia 81, 22–30 (2016).
Biezonski, D., Cha, J., Steinglass, J. & Posner, J. Evidence for thalamocortical circuit abnormalities and associated cognitive dysfunctions in underweight individuals with anorexia nervosa. Neuropsychopharmacology 41, 1560–1568 (2016).
Val-Laillet, D. et al. Neuroimaging and neuromodulation approaches to study eating behavior and prevent and treat eating disorders and obesity. NeuroImage. Clin. 8, 1–31 (2015).
Benjamin, L. & Wulfert, E. Dispositional correlates of addictive behaviors in college women: Binge eating and heavy drinking. Eat. Behav. 6, 197–209 (2005).
Jasinska, A. J. et al. Impulsivity and inhibitory control deficits are associated with unhealthy eating in young adults. Appetite 59, 738–747 (2012).
Oliva, R., Morys, F., Horstmann, A., Castiello, U. & Begliomini, C. The impulsive brain: Neural underpinnings of binge eating behavior in normal-weight adults. Appetite 136, 33–49 (2019).
Ochsner, K. N., Silvers, J. A. & Buhle, J. T. Functional imaging studies of emotion regulation: a synthetic review and evolving model of the cognitive control of emotion. Ann. N. Y. Acad. Sci. 1251, E1–E24 (2012).
Godet, A., Fortier, A., Bannier, E., Coquery, N. & Val-Laillet, D. Interactions between emotions and eating behaviors: Main issues, neuroimaging contributions, and innovative preventive or corrective strategies. Rev. Endocr. Metab. Disord. 23, 807–831 (2022).
Hollmann, M., Pleger, B., Villringer, A. & Horstmann, A. Brain imaging in the context of food perception and eating. Curr. Opin. Lipidol. 24, 18–24 (2013).
Shin, A. C., Zheng, H. & Berthoud, H. R. An expanded view of energy homeostasis: Neural integration of metabolic, cognitive, and emotional drives to eat. Physiol. Behav. 97, 572–580 (2009).
Rolls, E. Functions of the orbitofrontal and pregenual cingulate cortex in taste, olfaction, appetite and emotion. Acta Physiol. Hung. 95, 131–164 (2008).
Craig, A. D. B. How do you feel–now? The anterior insula and human awareness. Nat. Rev. Neurosci. 10, 59–70 (2009).
Jung, J. Y., Lambon Ralph, M. A. & Jackson, R. L. Subregions of DLPFC Display Graded yet Distinct Structural and Functional Connectivity. J. Neurosci. 42, 3241–3252 (2022).
Weygandt, M. et al. The role of neural impulse control mechanisms for dietary success in obesity. Neuroimage 83, 669–678 (2013).
Hollmann, M. et al. Neural correlates of the volitional regulation of the desire for food. Int. J. Obes. 36, 648–655 (2012).
Le, D. S. N. T. et al. Less activation of the left dorsolateral prefrontal cortex in response to a meal: a feature of obesity. Am. J. Clin. Nutr. 84, 725–731 (2006).
Baek, K., Morris, L. S., Kundu, P. & Voon, V. Disrupted resting-state brain network properties in obesity: Decreased global and putaminal cortico-striatal network efficiency. Psychol. Med. 47, 585–596 (2017).
Chen, F., He, Q., Han, Y., Zhang, Y. & Gao, X. Increased BOLD Signals in dlPFC Is Associated With Stronger Self-Control in Food-Related Decision-Making. Front. Psychiatry 9, 1–8 (2018).
Barth, B., Strehl, U., Fallgatter, A. J. & Ehlis, A.-C. Near-Infrared Spectroscopy based Neurofeedback of Prefrontal Cortex Activity: A Proof-of-Concept Study. Front. Hum. Neurosci. 10, 1–13 (2016).
Sitaram, R. et al. Closed-loop brain training: the science of neurofeedback. Nat. Rev. Neurosci. 18, 86–100 (2017).
Gaume, A., Vialatte, A., Mora-Sánchez, A., Ramdani, C. & Vialatte, F. B. A psychoengineering paradigm for the neurocognitive mechanisms of biofeedback and neurofeedback. Neurosci. Biobehav. Rev. 68, 891–910 (2016).
Thibault, R. T., MacPherson, A., Lifshitz, M., Roth, R. R. & Raz, A. Neurofeedback with fMRI: A critical systematic review. Neuroimage 172, 786–807 (2018).
Muñoz-Moldes, S. & Cleeremans, A. Delineating implicit and explicit processes in neurofeedback learning. Neurosci. Biobehav. Rev. 118, 681–688 (2020).
Autenrieth, M., Kober, S. E., Neuper, C. & Wood, G. How Much Do Strategy Reports Tell About the Outcomes of Neurofeedback Training? A Study on the Voluntary Up-Regulation of the Sensorimotor Rhythm. Front. Hum. Neurosci. 14, 1–12 (2020).
Kober, S. E., Witte, M., Ninaus, M., Neuper, C. & Wood, G. Learning to modulate one’s own brain activity: the effect of spontaneous mental strategies. Front. Hum. Neurosci. 7, 1–12 (2013).
Ros, T. et al. Consensus on the reporting and experimental design of clinical and cognitive-behavioural neurofeedback studies (CRED-nf checklist). Brain 143, 1674–1685 (2020).
Kadosh, K. C. & Staunton, G. A systematic review of the psychological factors that influence neurofeedback learning outcomes. Neuroimage 185, 545–555 (2019).
Pereira, D. J. et al. Neurofeedback-dependent influence of the ventral striatum using a working memory paradigm targeting the dorsolateral prefrontal cortex. Front. Behav. Neurosci. 17, 1–15 (2023).
Sherwood, M. S., Kane, J. H., Weisend, M. P. & Parker, J. G. Enhanced control of dorsolateral prefrontal cortex neurophysiology with real-time functional magnetic resonance imaging (rt-fMRI) neurofeedback training and working memory practice. Neuroimage 124, 214–223 (2016).
Kohl, S. H. et al. Real-time fMRI neurofeedback training to improve eating behavior by self-regulation of the dorsolateral prefrontal cortex: A randomized controlled trial in overweight and obese subjects. Neuroimage 191, 596–609 (2019).
Blume, F. et al. NIRS-based neurofeedback training in a virtual reality classroom for children with attention-deficit/hyperactivity disorder: study protocol for a randomized controlled trial. Trials 18, 41 (2017).
Marx, A.-M. et al. Near-infrared spectroscopy (NIRS) neurofeedback as a treatment for children with attention deficit hyperactivity disorder (ADHD)â€”a pilot study. Front. Hum. Neurosci. 8, 1–13 (2015).
Hudak, J. et al. Near-Infrared Spectroscopy-Based Frontal Lobe Neurofeedback Integrated in Virtual Reality Modulates Brain and Behavior in Highly Impulsive Adults. Front. Hum. Neurosci. 11, 1–13 (2017).
Kimmig, A.-C. S. et al. Feasibility of NIRS-based neurofeedback training in social anxiety disorder: behavioral and neural correlates. J. Neural Transm. 126, 1175–1185 (2019).
Aranyi, G., Pecune, F., Charles, F., Pelachaud, C. & Cavazza, M. Affective Interaction with a Virtual Character Through an fNIRS Brain-Computer Interface. Front. Comput. Neurosci. 10, 1–14 (2016).
Aranyi, G., Charles, F. & Cavazza, M. Anger-based BCI Using fNIRS Neurofeedback. in Proceedings of the 28th Annual ACM Symposium on User Interface Software & Technology 511–521 (ACM, 2015). doi:10.1145/2807442.2807447.
Yu, L. et al. Improving Emotion Regulation Through Real-Time Neurofeedback Training on the Right Dorsolateral Prefrontal Cortex: Evidence From Behavioral and Brain Network Analyses. Front. Hum. Neurosci. 15, 1–11 (2021).
Maywald, M. et al. Individual- and Connectivity-Based Real-Time fMRI Neurofeedback to Modulate Emotion-Related Brain Responses in Patients with Depression: A Pilot Study. Brain Sci. 12, 1714 (2022).
Weiss, F., Zhang, J., Aslan, A., Kirsch, P. & Gerchen, M. F. Feasibility of training the dorsolateral prefrontal-striatal network by real-time fMRI neurofeedback. Sci. Rep. 12, 1669 (2022).
Spetter, M. S. et al. Volitional regulation of brain responses to food stimuli in overweight and obese subjects: A real-time fMRI feedback study. Appetite 112, 188–195 (2017).
Strehl, U. What learning theories can teach us in designing neurofeedback treatments. Front. Hum. Neurosci. 8, 1–8 (2014).
Arns, M. et al. Neurofeedback: One of today’s techniques in psychiatry? Encephale. 43, 135–145 (2017).
Birbaumer, N., Ruiz, S. & Sitaram, R. Learned regulation of brain metabolism. Trends in Cognitive Sciences vol. 17 295–302 at https://doi.org/10.1016/j.tics.2013.04.009 (2013).
Ninaus, M. et al. Neural substrates of cognitive control under the belief of getting neurofeedback training. Front. Hum. Neurosci. 7, 1–10 (2013).
von Lühmann, A., Ortega-Martinez, A., Boas, D. A. & Yücel, M. A. Using the General Linear Model to Improve Performance in fNIRS Single Trial Analysis and Classification: A Perspective. Front. Hum. Neurosci. 14, 1–17 (2020).
Balleine, B. W. & O’Doherty, J. P. Human and rodent homologies in action control: Corticostriatal determinants of goal-directed and habitual action. Neuropsychopharmacology vol. 35 48–69 at https://doi.org/10.1038/npp.2009.131 (2010).
Delgado, M. R., Miller, M. M., Inati, S. & Phelps, E. A. An fMRI study of reward-related probability learning. Neuroimage 24, 862–873 (2005).
Balleine, B. W., Delgado, M. R. & Hikosaka, O. The Role of the Dorsal Striatum in Reward and Decision-Making: Fig. 1. J. Neurosci. 27, 8161–8165 (2007).
Zhao, Z. et al. Putamen volume predicts real-time fMRI neurofeedback learning success across paradigms and neurofeedback target regions. Hum. Brain Mapp. 42, 1879–1887 (2021).
Skottnik, L., Sorger, B., Kamp, T., Linden, D. & Goebel, R. Success and failure of controlling the real-time functional magnetic resonance imaging neurofeedback signal are reflected in the striatum. Brain Behav. 9, e01240 (2019).
Vogt, B. A. Cingulate cortex in the three limbic subsystems. in Handbook of Clinical Neurology vol. 166 39–51 (Elsevier B.V., 2019).
Rolls, E. T. The cingulate cortex and limbic systems for emotion, action, and memory. Brain Struct. Funct. 224, 3001–3018 (2019).
Hardwick, R. M., Caspers, S., Eickhoff, S. B. & Swinnen, S. P. Neural correlates of action: Comparing meta-analyses of imagery, observation, and execution. Neuroscience and Biobehavioral Reviews vol. 94 31–44 at https://doi.org/10.1016/j.neubiorev.2018.08.003 (2018).
Liakakis, G., Nickel, J. & Seitz, R. J. Diversity of the inferior frontal gyrus—A meta-analysis of neuroimaging studies. Behav. Brain Res. 225, 341–347 (2011).
Hagoort, P. On Broca, brain, and binding: a new framework. Trends Cogn. Sci. 9, 416–423 (2005).
Bookheimer, S. Functional MRI of Language: New Approaches to Understanding the Cortical Organization of Semantic Processing. Annu. Rev. Neurosci. 25, 151–188 (2002).
Morin, A. & Michaud, J. Self-awareness and the left inferior frontal gyrus: Inner speech use during self-related processing. Brain Res. Bull. 74, 387–396 (2007).
Emmert, K. et al. Meta-analysis of real-time fMRI neurofeedback studies using individual participant data: How is brain regulation mediated? Neuroimage 124, 806–812 (2016).
Pearson, J. M., Heilbronner, S. R., Barack, D. L., Hayden, B. Y. & Platt, M. L. Posterior cingulate cortex: Adapting behavior to a changing world. Trends in Cognitive Sciences vol. 15 143–151 at https://doi.org/10.1016/j.tics.2011.02.002 (2011).
Leech, R. & Sharp, D. J. The role of the posterior cingulate cortex in cognition and disease. Brain vol. 137 12–32 at https://doi.org/10.1093/brain/awt162 (2014).
Phan, K. L., Wager, T., Taylor, S. F. & Liberzon, I. Functional neuroanatomy of emotion: A meta-analysis of emotion activation studies in PET and fMRI. Neuroimage 16, 331–348 (2002).
Etkin, A., Büchel, C. & Gross, J. J. The neural bases of emotion regulation. Nat. Rev. Neurosci. 16, 693–700 (2015).
Helion, C., Krueger, S. M. & Ochsner, K. N. Emotion regulation across the life span. Handb. Clin. Neurol. 163, 257–280 (2019).
Krylova, M. et al. Progressive modulation of resting-state brain activity during neurofeedback of positive-social emotion regulation networks. Sci. Rep. 11, 23363 (2021).
Sorger, B., Scharnowski, F., Linden, D. E. J., Hampson, M. & Young, K. D. Control freaks: Towards optimal selection of control conditions for fMRI neurofeedback studies. Neuroimage 186, 256–265 (2019).
Fede, S. J., Dean, S. F., Manuweera, T. & Momenan, R. A Guide to Literature Informed Decisions in the Design of Real Time fMRI Neurofeedback Studies: A Systematic Review. Front. Hum. Neurosci. 14, (2020).
Kohl, S. H. et al. The Potential of Functional Near-Infrared Spectroscopy-Based Neurofeedback—A Systematic Review and Recommendations for Best Practice. Front. Neurosci. 14, (2020).
Peirce, J. et al. PsychoPy2: Experiments in behavior made easy. Behav. Res. Methods 51, 195–203 (2019).
Delpy, D. T. et al. Estimation of optical pathlength through tissue from direct time of flight measurement. Phys. Med. Biol. 33, 1433–1442 (1988). Godet, A., Serrand, Y., Fortier, A., Léger, B., Bannier, E., Val-Laillet, D., & Coquery, N. (2023). Subjective feeling of control during fNIRS-based neurofeedback targeting the DL-PFC is related to neural activation determined with short-channel correction. POS ONE in minor revision.
Gramfort, A. et al. MEG and EEG data analysis with MNE-Python. Front. Neurosci. 7, 1–13 (2013).
Abraham, A. et al. Machine learning for neuroimaging with scikit-learn. Front. Neuroinform. 8, 1–10 (2014).
Pollonini, L., Bortfeld, H. & Oghalai, J. S. PHOEBE: a method for real time mapping of optodes-scalp coupling in functional near-infrared spectroscopy. Biomed. Opt. Express 7, 5104 (2016).
Cardenas, V. A. et al. HHS Public Access. 15, 63–72 (2022).
Coquery, N. et al. Brain Responses to Food Choices and Decisions Depend on Individual Hedonic Profiles and Eating Habits in Healthy Young Women. Front. Nutr. 9, (2022).
Nieto-Castanon, A. Handbook of functional connectivity Magnetic Resonance Imaging methods in CONN. Handbook of functional connectivity Magnetic Resonance Imaging methods in CONN (Hilbert Press, 2020). doi:10.56441/hilbertpress.2207.6598.
Stopyra, M. A. et al. Altered functional connectivity in binge eating disorder and bulimia nervosa: A resting-state fMRI study. Brain Behav. 9, 1–13 (2019).
Zalesky, A., Fornito, A. & Bullmore, E. T. Network-based statistic: Identifying differences in brain networks. Neuroimage 53, 1197–1207 (2010).
Zalesky, A., Cocchi, L., Fornito, A., Murray, M. M. & Bullmore, E. Connectivity differences in brain networks. Neuroimage 60, 1055–1062 (2012).
Zeger, S. L. & Liang, K.-Y. Longitudinal Data Analysis for Discrete and Continuous Outcomes. Biometrics 42, 121 (1986).
Zeger, S. L., Liang, K.-Y. & Albert, P. S. Models for Longitudinal Data: A Generalized Estimating Equation Approach. Biometrics 44, 1049 (1988).

Table 1. Results of the univariate and multivariate GEE analysis on the subjective gauge control feeling.

Control as dependant variable	UNIVARIATE			MULTIVARIATE
Variables	B	CI 95%	p-value	B	IC 95%	p-value
Age	0.088	[-0.121-0.297]	0.407
Hunger	-0.177	[-0.340-0.014]	0.034	-0.171	[-0.326 - -0.017]	0.030
Thirst	-0.75	[-0.379-0.229]	0.628
Well-being	-0.048	[-0.319-0.223]	0.727
Motivation	0.277	[0.031-0.524]	0.028	0.231	[0.002-0.406]	0.048
Effort_intensity	0.020	[-0.393-0.434]	0.924

Note: p-value threshold, p<0.05; CI, confidence interval.

Table 2. Results of the second-level analysis of the 1^st and 8^th NF sessions (n=30) on HbO concentration.

Sources	Regions	Beta	Std E.	Z score	P>\|z\|	[0,025	0,975]	p-bonf
S1_D1	dlPFC	1.774	0.977	1.816	0.069	-0.141	3.689	1
S1_D2	dlPFC	10.472	1.225	8.546	0.000	8.070	12.873	<0.001
S2_D3	motor	2.522	1.690	1.492	0.136	-0.790	5.834	1
S3_D1	OthFC	0.569	0.907	0.628	0.530	-1.208	2.346	1
S3_D4	OthFC	-0.528	0.907	-0.582	0.561	-2.305	1.249	1
S4_D2	OthFC	10.091	1.190	8.483	0.000	7.759	12.422	<0.001
S4_D4	OthFC	-0.413	1.023	-0.403	0.687	-2.419	1.593	1
S4_D5	OthFC	5.701	0.991	5.750	0.000	3.758	7.644	<0.001
S5_D4	OthFC	-1.005	0.899	-1.118	0.264	-2.767	0.757	1
S5_D6	OthFC	-0.063	0.899	-0.070	0.944	-1.825	1.699	1
S6_D4	OthFC	-0.175	0.916	-0.191	0.848	-1.970	1.620	1
S6_D6	OthFC	0.448	0.899	0.498	0.618	-1.314	2.210	1
S6_D7	OthFC	0.338	0.915	0.370	0.712	-1.455	2.131	1
S7_D5	dlPFC	4.967	1.130	4.397	0.000	2.753	7.182	<0.001
S7_D7	dlPFC	0.135	0.944	0.143	0.886	-1.715	1.984	1
S8_D6	OthFC	1.512	0.899	1.682	0.093	-0.250	3.274	1
S8_D7	OthFC	0.500	0.907	0.552	0.581	-1.277	2.277	1

Abbreviations: dlPFC, dorsolateral prefrontal cortex; OthFC, other frontal channels; Std E., standard error; p-bonf, p-value with Bonferroni correction <0.05.

Table 3. fMRI brain activation during NF sessions. Based on SVC analysis, only brain activations that survived FWE correction (p<0.05) for multiple comparisons at the cluster level and at the voxel level with puncorr.<0.001 within the dlPFC (our a priori hypothesis) and within the ROI identified in the activated clusters, are reported. All ROIs were statistically adjusted for age. The table describes extent (number of voxels, k), right or left hemisphere (L, R), Brodmann areas (BA), T-values and p-values for the statistical contrast, Montreal Neurological Institute (MNI) coordinates and T and p-values for the peak. Brain regions are listed by extent (K) and L/R hemisphere.

	Anatomical area			p-value cluster			MNI peak coordinates			p-value peak
K	Structure	L/R	BA	FWE	FDR	Uncorr.	x	y	z	T	Uncorr.
574	PCG	L	6	<0.001	<0.001	<0.001	-40	-5	45	7.88	<0.001
380	PCG	R	6	<0.001	<0.001	<0.001	43	1	38	8.71	<0.001
479	SMA	L	6,8,32	<0.001	<0.001	<0.001	-7	6	53	6.84	<0.001
460	SMA	R	6,8,32	<0.001	<0.001	<0.001	8	1	53	6.80	<0.001
441	Insula	L	48	<0.001	<0.001	<0.001	-30	18	5	7.33	<0.001
407	Insula	R	48	<0.001	<0.001	<0.001	33	26	-3	8.60	<0.001
149	SMG	L	2,40,48	<0.001	0.002	0.002	-60	-45	28	6.20	<0.001
410	SMG	R	2,40,48	<0.001	<0.001	<0.001	63	-37	40	6.97	<0.001
227	IFGoper	L	44,48	<0.001	<0.001	<0.001	-50	11	0	6.42	<0.001
201	Put	L		<0.001	<0.001	<0.001	-25	13	8	5.38	<0.001
165	Put	R		<0.001	0.001	0.001	33	13	-5	5.61	<0.001
140	IFGtri	L	45	<0.001	0.013	0.002	-37	21	5	6.24	<0.001
120	IFGtri	R	45	<0.001	0.022	0.004	46	23	-3	6.66	<0.001
91	IFGorb	L	47	<0.001	0.022	0.011	-37	28	-5	5.58	<0.001
111	IFGorb	R	47	<0.001	0.012	0.006	38	26	-8	7.43	<0.001
131	MCC	L	32	<0.001	0.003	0.003	1	13	43	6.12	<0.001
151	MCC	R	32	<0.001	0.002	0.008	6	16	35	6.59	<0.001
112	dlPFC	L	9,46	<0.001	<0.001	0.006	-40	48	23	5.03	<0.001
93	dlPFC	R	9,46	<0.001	<0.001	0.010	43	38	28	5.36	<0.001
91	Pal	L		<0.001	0.011	0.011	-15	1	-5	4.93	<0.001
30	Pal	R		0.004	0.232	0.116	21	3	5	4.35	<0.001
28	ACC	L	24,32	0.008	0.128	0.128	-10	23	28	4.94	<0.001
7	ACC	R	24,32	0.024	0.442	0.442	8	18	28	4.21	<0.001
20	Thal VA	L		0.003	0.193	0.193	-15	-2	8	3.96	<0.001
12	Thal VA	R		0.004	0.621	0.311	13	1	5	3.89	<0.001
3	Thal VL	L		0.021	0.800	0.629	-15	-7	13	3.71	<0.001
8	Thal VL	R		0.013	0.709	0.410	13	13	3	3.67	<0.001
13	Cau	L		0.024	0.583	0.291	-15	-2	15	3.65	<0.001
16	Cau	R		0.021	0.364	0.243	18	-2	15	4.15	<0.001

Abbreviations: ACC, Anterior Cingulate Cortex; Cau, caudate nucleus; IFGoper, Inferior Frontal Gyrus Opercular part, Inferior Frontal Gyrus pars orbitalis; IFGtri, Inferior Frontal Gyrus triangular part, IFGober; Ins, insula; MCC, Middle Cingulate Cortex; Pal, Pallidum; PGC, Precentral Gyrus; Put, Putamen; SGM, Supramarginal Gyrus; SMA, Supplementary Motor Area; Thal VA, ventral anterior Thalamus; Thal VL, ventral lateral Thalamus; . SVC, small volume correction; FWE, family-wise error.

Table 4. FC analysis, clusters of ROI-to-ROI connections based on the SPC approach.

Analysis Unit	Statistic	p-unc	p-FDR	p-FWE
One-sided: positive
Cluster 1/2	Score = 36.43	0.010	0.019	0.012
	Mass = 119.81	0.002	0.005	0.002
	Size = 12	0.001	0.002	0.001
Connection PCC - R Put	T(13) = 3.62	0.002	0.047
Connection L dlPFC - PCC	T(13) = 3.47	0.002	0.047
Connection R dlPFC - PCC	T(13) = 3.24	0.003	0.049
Connection PCC - R Cau	T(13) = 3.04	0.005	0.054
Connection PCC - L Put	T(13) = 2.76	0.008	0.060
Connection PCC - L Cau	T(13) = 2.74	0.009	0.060
Analysis Unit	Statistic	p-unc	p-FDR	p-FWE
One-sided: negative
Cluster 1/1	Score = 58.63	0.004	0.004	0.004
	Mass = 49.83	0.023	0.023	0.029
	Size = 2	0.254	0.254	0.254
Connection L Ins - L dlPFC	T(13) = -4.99	<0.001	0.006

Abbreviations: Cau, caudate; dlPFC, dorsolateral prefrontal cortex; PCC, posterior cingulate cortex; Put, putamen; L, left; R, right; p-unc, uncorrected; FDR, false-discovery rate; FWE, family-error wise.

No competing interests reported.

Download PDF

Journal Publication

published 28 Aug, 2024

Read the published version in Scientific Reports →

Editorial decision: Revision requested
26 Mar, 2024
Reviews received at journal
11 Mar, 2024
Reviewers agreed at journal
25 Feb, 2024
Reviewers agreed at journal
02 Oct, 2023
Reviewers invited by journal
14 Sep, 2023
Editor assigned by journal
14 Sep, 2023
Editor invited by journal
03 Aug, 2023
Submission checks completed at journal
03 Aug, 2023
First submitted to journal
06 Jul, 2023

You are reading this latest preprint version

Functional near-infrared spectroscopy-based neurofeedback training induces changes in cortico-striatal functional connectivity: A proof of concept study

Status:

Journal Publication

Version 1

Abstract

Figures

INTRODUCTION

RESULTS

DISCUSSION

Limits and overall conclusion

METHODS

Ethics statement

Participants

Study procedure

Self-report questionnaires

Neurofeedback task

Functional near-infrared spectroscopy (fNIRS) system

MRI data acquisition

Statistics

Declarations

References

Tables

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1