Abstract
We urgently need new therapeutic strategies for depression (1). Depression is one of the top three causes of disability in the global disease burden statistic, affecting up to 15% of the population of high-income countries and with increasing prevalence also in low- and middle-income countries. This comes at huge socioeconomic and healthcare costs, especially because a large number of patients develop chronic illness, regardless of the available treatments that are effective for the majority of patients. The mainstay of current management are pharmacological and psychological/psychosocial interventions, and recent innovation has been particularly active in the field of physical interventions, adding transcranial magnetic stimulation (TMS) to the repertoire.
12.1 Introduction
We urgently need new therapeutic strategies for depression (1). Depression is one of the top three causes of disability in the global disease burden statistic, affecting up to 15% of the population of high-income countries and with increasing prevalence also in low- and middle-income countries. This comes at huge socioeconomic and healthcare costs, especially because a large number of patients develop chronic illness, regardless of the available treatments that are effective for the majority of patients. The mainstay of current management are pharmacological and psychological/psychosocial interventions, and recent innovation has been particularly active in the field of physical interventions, adding transcranial magnetic stimulation (TMS) to the repertoire. Limitations of current treatment options include medication side effects, nonresponse, and frequent relapse. The scale of the public health problem and the limitations of existing treatments underscore the need for better, and more effective, treatment and relapse prevention options. In our opinion, interventions that involve the active collaboration of the patient are particularly promising, which is why the neurofeedback approach that has seen a resurgence in recent years is conceptually rather attractive.
Since its invention over twenty-five years ago, functional magnetic resonance imaging (fMRI) has become one of the most widely used and publicly visible noninvasive techniques to measure brain activation. fMRI-based neurofeedback (fMRI-NF) has the potential to open up new paths to translation. During fMRI-NF training, participants receive feedback on their brain activity in real-time and are instructed to change this activation, for example, by engaging in specific mental imagery. One attractive feature of neurofeedback is that it enables patients to control their own brain activity and thus contributes to their experience of self-efficacy, which is an important therapeutic factor in many neuropsychiatric disorders.
Recent advances in affective neuroscience in general and its application to depression, reviewed in Chapters 7 and 8 of this book, have paved the ground for the identification of neurofeedback targets (2). Modulation of prefrontal cortex and limbic areas could be used to improve emotion regulation, modulation of amygdala, insula and other parts of the salience network to normalize emotional reactivity, modulation of frontoparietal circuits or the default mode network (DMN) to attenuate rumination and tackle cognitive symptoms of depression, and modulation of the reward system to address anhedonia and the amotivational syndrome. The syndromal, multifaceted nature of depression poses a challenge to any unified treatment approach, but also plenty of opportunities to target specific neural substrates with neurofeedback. It is thus perhaps not surprising that depression is one of the clinical areas where fMRI-NF research has advanced most.
12.2 The Neural Basis of Neurofeedback
Although neurofeedback, particularly with EEG, has been applied for decades in research and in clinical settings (3), relatively little is known about the neural effects of neurofeedback. For neurofeedback guided self-regulation, previous research suggests an interplay of reward processing, self-regulation, and learning mechanisms in interaction with brain networks involved in the specific mental task driving the feedback (13). However, studies that investigated the general neural mechanisms of neurofeedback on the whole-brain level are sparse.
Notably, a recent meta-analysis compared whole-brain activation across different neurofeedback tasks, and thereby revealed extensive overlap in brain activation across neurofeedback studies in prefrontal, parietal, occipital as well as subcortical areas and deactivation of the DMN (4). The observed network appeared to be congruent with the main theorized psychological components of neurofeedback interventions (Figure 12.1) and included regions recruited in self-regulation and executive control, particularly the ventrolateral and dorsolateral prefrontal cortex (vlPFC and dlPFC), the anterior cingulate cortex (ACC), the anterior insula (aINS), and clusters in the parietal cortex. Furthermore, it involved activation in regions involved in visual feedback processing and learning, such as the occipital cortex, the basal ganglia (notably the dorsal and ventral striatum), and the thalamus. In addition, deactivation was observed for main hubs of the DMN, precuneus, posterior cingulate cortex, and lateral parietal cortex, as well as deactivation in Heschl’s gyrus, potentially reflecting attention shifts away from auditory processing of scanner noise.
Figure 12.1 Regions recruited during neurofeedback guided self-regulation. A. A distributed network of regions implicated in cognitive control is activated during neurofeedback including lateral parietal and medial as well as lateral prefrontal areas. B. The default mode network shows modulations during neurofeedback that are likely due to the task demand and shifts in internally and externally directed attention, also reflected by deactivation of auditory areas in the lateral temporal lobe. C. Regions implicated in reward learning and visual processing are reliably recruited, including the visual cortex, the anterior cingulate cortex, the anterior insula, the basal ganglia and the thalamus. Abbreviations: ACC: Anterior cingulate cortex; DLPFC: dorsolateral prefrontal cortex; VLPFC: ventrolateral PFC; LPL: lateral parietal lobe; aINS: Anterior insula; PCC/PreC: posterior cingulate cortex / precuneus; MPFC: Medial prefrontal cortex; AG: Angular gyrus; LTL: lateral temporal lobe; TAL: Thalamus; BG: basal ganglia; VC: visual cortex.
By comparing activation across different (affective as well as non-affective) neurofeedback tasks, the study by Emmert et al. (4) revealed a network of regions that is generally recruited during neurofeedback, but not necessarily specific for neurofeedback. Increases in the implicated parietal-prefrontal regions are also observed during various self-regulation tasks without neurofeedback (5–7), and deactivation of the DMN is reliably associated with attention-demanding tasks (8–10).
In addition to these shared neural components across neurofeedback tasks, distinct mental tasks and neural targets affect distinguishable, task-specific networks (11). Yet so far, no large-scale comparisons of multiple neurofeedback paradigms have been made, so it is not clear whether certain subgroups of neurofeedback approaches share an even more pronounced neural basis.
In addition to the scarceness of comprehensive whole-brain analysis on neurofeedback, the field also lacks evidence on the temporal properties of the neural processes occurring within involved networks. Taking into account how crucial timing is in operant conditioning (95, 96), these questions appear to be fundamental for the understanding how neurofeedback training can induce learning and reshape the brain. In one of our recent studies, we aimed to contribute to this issue by analyzing brain action across different self-regulation tasks, with and without providing neurofeedback (12). Self-regulation with feedback was accompanied by stronger activation in the striatum, and additional time-resolved analysis revealed that neurofeedback performance was positively correlated with a delayed brain response in the striatum that reflected the accuracy of self-regulation.
Overall, the current state of research suggests that, during neurofeedback interventions, task-general self-regulation processes execute control on mental task-specific areas beyond the neurofeedback target region. During this process, successful self-regulation performance is reinforced through positive feedback. For neurofeedback as a clinical tool, it remains to be specified to what extent processes specific to a given neurofeedback intervention and unspecific effects, such as reinforcement of general self-regulation abilities or improved self-efficacy, differentially contribute to treatment outcomes.
12.3 fMRI-NF Neurofeedback Treatments for Depression
While neurofeedback approaches differ with regard to the neural target that participants train to control, as well as the mental processes used to control the neurofeedback signal, previous theoretical accounts of neurofeedback have argued that neurofeedback-guided self-regulation generally implicates three main components (13): general self-regulation, reward learning, and processes specific to the self-regulation task.
12.3.1 Mechanisms Shared across fMRI-NF Neurofeedback Treatments for Depression
12.3.1.1 Self-regulation
The network recruited during neurofeedback across neurofeedback tasks includes areas implicated in self-regulation across various cognitive and affective tasks (4, 12). Of the recruited regions, especially the aINS, vlPFC, dlPFC, and ACC have previously also been related to different forms of top-down control in emotion regulation: particularly the aINS, vlPFC, dlPFC, and ACC have been shown to contribute to the endogenous generation of emotional states of positive as well as of negative valence, across different self-regulation modalities (5). Additionally, they are recruited during downregulation of negative emotions across various emotion regulation strategies (6),
The task-unspecific recruitment of these areas in self-regulation suggests that a self-regulation network that contributes to cognitive control in various mental domains is reinforced across different neurofeedback approaches. Our recent neurofeedback study (14) supports this notion also in depression: a neurofeedback control group that performed primarily non-affective self-regulation (visual scenes imagery) showed significant improvements in clinical symptoms that were comparable to improvements of the (emotion-regulation) intervention group. These effects exceeded the expected improvements of placebo effects of other high-tech interventions in depression significantly. Notably, placebo-controlled neurofeedback trials on depression, in which self-regulation performance was not matched between intervention and control group, did not show corresponding improvements for the control group (15). Taken together, these results suggest that neurofeedback regulation alters symptoms of depression across specific neurofeedback tasks, but it is not clear what causes such general effects of neurofeedback. On the one hand, they could be related to improvements in general self-regulation abilities, but on the other hand they could also be related to unspecific effects of positive feedback or increases in self-efficacy.
In addition to top-down control, self-observation constitutes an intrinsic feature of mental self-regulation tasks that supports successful self-regulation (16, 17). It is therefore likely that introspective abilities contribute to such domain-general effects. Of the regions recruited during neurofeedback, especially the anterior insula and the ACC have been shown to play a selective role in introspection (18). Additionally, several studies support a link between altered insula and ACC functioning and alexithymia (19–22).
In the presence of pronounced deficits of subjective experience of internal states, the neurofeedback signal could constitute an external information source on ongoing mental activity. Notably, Zotev et al. (23) were able to show that neurofeedback performance during emotional memory recall was negatively correlated with alexithymia ratings in healthy participants, suggesting a relationship between perception of internal states and neurofeedback performance.
A recent neurofeedback approach motivated by this property of neurofeedback has provided depressed patients with neurofeedback on the effectiveness of mental strategies to control ACC reactivity to negative affective content in depression (24). Neurofeedback performance could predict whether strategies were experienced as being difficult to perform and efficient for controlling negative mood during the neurofeedback training, but were unrelated to ratings acquired before the training, suggesting that the information provided by neurofeedback was indeed used to evaluate subjective experiences. After a one-month follow-up, neurofeedback performance remained predictive of efficacy ratings and predicted how often patients would use certain self-regulation strategies in daily life.
Self-efficacy could be another crucial factor contributing to task-general effects, as perceived self-efficacy shows a negative correlation with subclinical depressive symptoms (25). Additionally, Kavanagh and Wilson (26) showed that improvements in depression correlated with self-efficacy of mood regulation in cognitive therapy and could be used to identify patients who showed remission over the following twelve months (see also Maddux and Meier (27)).
12.3.1.2 Reinforcement and Regulation of Neural States
When providing feedback on self-regulation performance, successful self-regulation is accompanied by increased activation in the striatum (12), a key region of reward learning (28, 29). It has been shown that neurofeedback with patients with depression can lead to increased coupling between the neurofeedback target area and the dorsal and ventral striatum (30), suggesting that self-regulation reinforcement takes place during depression treatment with neurofeedback.
Besides reinforcing top-down self-regulation, another property of neurofeedback is its association with a decreased activation in the DMN (4, 12). Notably, it has been shown that hyperactivity of the DMN contributes to impaired self-regulation in depression (31) and modulations in DMN connectivity have been related to increased rumination in depression (32, 33). Taking into account the well-observed anticorrelation between attention networks and the DMN, reinforcing an upstate in executive networks as well as a downstate of the DMN could help to reduce the distorting influence of the DMN on ongoing processing in depression (34, 35).
In addition to being associated with a general decrease in DMN activation, neurofeedback provides the possibility to alter specific configurations of DMN connectivity: Young et al. (30) showed that neurofeedback-guided self-regulation with autobiographical memories altered connectivity between the amygdala and various nodes of the DMN. Connectivity changes were associated with memory recall and translated to post-scan resting-state measures. As DMN connectivity has repeatedly been related to alterations in self-referential thought in depression (31–33), Young and colleagues argued that neurofeedback could help to modulate distorted connectivity pattern related to negative self-referential rumination.
Besides its relationship to the DMN, rumination in depression has been repeatedly related to alterations in limbic activity (36–39). Previous neurofeedback approaches have used limbic regions as neurofeedback targets, either for modulating activation in relation to positive (15, 40, 41) or to negative valence (24, 42). The results of Young et al. (15) revealed that neurofeedback can alter memory recall of autobiographical affective content, a crucial factor contributing to excessive rumination in depression (43, 44).
Overall, reinforcement learning thereby suggests the strong possibility of neurofeedback to modulate automatic neural processes in depression that are not directly accessible for cognitive control. While other forms of self-regulation, such as cognitive reappraisal or meditation, rely on voluntary self-regulation, neurofeedback can even reinforce neural target states when participants are unaware of receiving neurofeedback (45). However, the exact effects of this reinforcement learning likely differ between different neurofeedback target and self-regulation strategies.
12.3.2 Different fMRI-NF Neurofeedback Approaches in Depression
While different neurofeedback approaches share basic mechanisms of self-regulation and reinforcement learning, existing fMRI-NF neurofeedback approaches show considerable variance in methodology (for a recent overview see Thibault et al. (46)). In depression, basic differences between neurofeedback approaches exist with regard to the targeted psychological mechanisms, as well with regard to which neural markers were selected to create the neurofeedback signal (see Tables 12.1 and 12.2)
| Year Published | Authors | Target process | Neural target | Group design | n | Neurofeedback regulation | Clinical improvements | Transfer | Follow-up | Neural outcome measure | |||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Baseline | Time | Group | Time | Group | |||||||||
| 2012 | Linden et al. |
| Individual areas responsive to positive affect | NFI, SR | [8 8] | Y | N | – | Y | Y | – | – | NFP |
| 2014 | Young et al. |
| Amygdala | NFI, OB | [14 7] | Y | Y | Y | – | – | Y | – | NFP |
| 2014 | Yuan et al. |
| Amygdala | NFI, OB, H | [14 13 27] | – | – | – | Y | N | Y | Y | Connectivity |
| 2016 | Zotev et al. |
| Amygdala | NFI, OB | [13 11] | – | – | – | – | – | Y | – | Correlation with EEG |
| 2016 | Hamilton et al. | Salience of negative stimuli: downregulation | Individual salience network region | NFI, Sham | [10 10] | – | – | N | – | – | Y | – | Negative reactivity |
| 2017 | Yamada et al. | – | FC (increase)between left DLPFC and left precuneus/PCC | – | [3] | – | (Y) | – | (N) | – | – | – | NFP |
| 2017a | Young et al. |
| Amygdala | NFI, OB | [19 17] | Y | Y | Y | Y | Y | – | Y | NFP |
| 2017b | Young et al. |
| Amygdala | NFI, OB | [18 16] | – | – | – | – | – | Y | Y | ROI activation |
| 2018 | Young et al. |
| Amygdala | NFI, OB | [18 16] | – | – | – | – | – | Y | Y | Connectivity |
| 2018 | MacDuffie | Reactivity to negative stimuli: downregulation | ACC | NFI | [13] | – | – | – | – | – | – | Y | Correlation with NFP |
| 2018 | Mehler et al. |
| Individual areas responsive to positive affect | NFI, OB | [16 16] | Y | Y | N | Y | N | N | Y | NFP |
| Year registered | Initiator | Target process | Neural target | Group design | Transfer | Follow-up | Main neural outcome measure | Identifier |
|---|---|---|---|---|---|---|---|---|
| 2013 | Moll | Blame related memories | Connecivity: anterior temporal lobe with septal/subgenual cingulate | NFI, OC | – | – | Connectivity/NFP | NCT01920490 |
| 2016 | Young | Positive memories: upregulation | Amygdala | NFI, OB | – | Y | – | NCT02709161 |
| 2016 | Peciña | Positive mood induction | rACC | Placebo, Medicated | – | – | NFP | NCT02674529 |
| 2017 | Scharnowski | Not defined | Not defined |
| – | Y | WBA | NCT03165578 |
| 2017 | Mathiak | Self-regulation abilities | PFC | NFI, OB [Depression Schizophrenia]; H | – | Y | NFP | NCT03183947 |
| 2018 | Young | Positive memories: upregulation | Amygdala | NFI, OB | – | – | NFP | NCT03428828 |
Abbreviations: Y = yes, N = no; NFI = neurofeedback intervention, OB = other brain region, H = healthy participants, SR = self-regulation without NF, OC =other connectivity marker; NFP = neurofeedback performance; WBA = whole-brain activation. Retrieved from ClinicalTrials.gov.
12.3.2.1 Emotion Regulation: Positive Affect
General Positive Affect
Our first fMRI-NF neurofeedback study performed with depressed patients focused on increasing activation in regions related to positive reactivity, without selecting neurofeedback target regions a priori (40). Individualized target regions were instead selected based on activation in response to positive images during a functional localizer scan. Taking into account that affective states show considerable interindividual variability with regard to associated brain activation (47), this approach ensured that regions were selected that were maximally responsive to positive experience in each participant. Concerning the content of self-regulation strategies, participants were free to modulate activation using individual emotion regulation strategies related to positive affect. Thereby this neurofeedback procedure aimed at training individually sensitive aspects of positive affect, without restricting self-regulation a priori to an affective subcomponent such as salience or hedonistic value. Results indicated that participants were able to increase activation in the individual regions of interest (ROIs) using positive emotion regulation. In comparison to a control group that performed emotion regulation without receiving neurofeedback, depressive symptoms significantly improved.
While these findings provided first evidence for the clinical relevance of fMRI-NF neurofeedback in depression, they were obtained through a small, non-randomized study and not controlled for unspecific effects of neurofeedback, for example, the placebo effect caused by exposure to a high-technology treatment environment (as has been described in response to sham TMS in depression, see Berlim et al. (48), Berlim et al. (49)). Recently, our group addressed this issue in a randomized clinical trial (14). This trial compared the approach described by Linden et al. (40) to a neurofeedback control protocol that trained patients to increase activation in a non-affective, visual imagery task, using the parahippocampal place area as target region. Although no significant difference between groups was found, there was significant pre–post improvement in depression scores for both groups beyond expected placebo effects, suggesting that a clinically relevant mechanism may have been modulated in both neurofeedback groups (see Section 12.3.1).
Saliency of Positive Affective Experiences
Instead of aiming at generally increasing positive affect, the neurofeedback approach by Young et al. (50) focused on modulating a specific subcomponent of positive affect, that is, the salience of positive affective experience. In depression, salience responses to stimuli with positive valence are significantly impaired (51). In order to improve saliency of positive information in depressed patients, Young et al. (50) provided participants with amygdala neurofeedback that they trained to increase by contemplating positive autobiographical memories. The choice of the amygdala as neurofeedback target was motivated by its multifaceted relevance in depression: in comparison to healthy individuals, patients with depression show increased reactivity to negative stimuli (52–54) and attenuated reactivity to positive stimuli (55, 56). Furthermore, it has been shown that the amygdala is a central node of the salience network (57, 58) and modulates the memory system based on affective arousal (59).
Upregulation of amygdala activity using positive autobiographical memories appeared to be effective for reducing clinical symptoms of depression in comparison to a control group receiving neurofeedback from a task-unrelated brain region. Whole-brain activation during a transfer run indicated increased activation in the temporal pole, superior temporal gyrus, and the thalamus for the experimental group in comparison to the control group. These structures have shown to be crucially involved in autobiographical memory (60–62), suggesting connectivity alterations specific to the trained mental task.
The training effects appeared to extend to post intervention mood ratings, with amygdala neurofeedback being associated with improved mood in indices of positive as well as to negative valence (50), supporting the effectiveness of amygdala-focused treatments for emotional states with positive as well as negative valence. A second clinical trial replicated the clinical improvements of this neurofeedback approach with higher sample size (15) and, currently, two ongoing clinical trials further test clinical efficacy of this approach by examining whether neurofeedback can support cognitive-behavioral therapy (63) and by targeting treatment-unresponsive patients (64), see Table 12.2.
In addition to outcomes in primary clinical measures, more general (neural) intervention effects were further investigated by Yuan et al. (41): comparison between pre- and post-training resting-state scans revealed elevated hypoconnectivity of the amygdala after the training, which predicted decreases in depression severity for the intervention group. Specifically, alterations in connectivity between the amygdala, temporal regions, and the hippocampus were observed, supporting mental task-specific alterations in the memory system. Connectivity analysis of the second trial data set (30) underlined the relationship between alterations in amygdala connectivity with training outcomes: Amygdala connectivity to the precuneus and the inferior frontal gyrus during neurofeedback predicted symptom improvements, suggesting clinically relevant alterations in processing of self-referential information and emotion regulation (for self-referential processing related to the precuneus, see Zhu et al. (32), Hamilton et al. (33), Sheline et al. (31) and for IFG (inferior frontal gyrus) involvement in positive emotion-regulation, see Engen et al. (5), Engen et al. (65)).
Notably, Young et al. (66) additionally demonstrated that effects of this neurofeedback approach could transfer to amygdala reactivity beyond the neurofeedback training. Increased amygdala activation during neurofeedback was associated with increased amygdala reactivity to happy faces and decreased reactivity to sad faces, as well as improved processing of positive stimuli in a behavioral test battery. Such transfer from neurofeedback training runs to markers of emotional reactivity provides a promising outlook for neurofeedback as a therapeutic tool, as this suggests that neurofeedback can induce changes in bottom-up-driven processes in depression.
12.3.2.2 Emotion Regulation: Negative Affect
General Negative Affect
So far, extensive research that focuses on self-regulation of general negative affect with fMRI-NF neurofeedback is lacking for depression. A previous paradigm developed by MacDuffie et al. (24) has, however, used neurofeedback to evaluate effectiveness of CBT strategies to downregulate ACC activation in response to negative autobiographical content. As self-regulation behavior was significantly predicted by neurofeedback performance even one month after the neurofeedback intervention, results demonstrate the strong relevance of self-regulation of brain activation related to negative affective states for the treatment of depression. However, despite applying a functional localizer, target areas for this study were restricted to the ACC and results thereby likely reflect a preselection of affective processing that involves the ACC. Additionally, a currently running clinical trial (Mathiak (67), see Table 12.2) aims to train participants to regulate PFC activation using emotional reappraisal, a commonly used, effective strategy for regulation of negative affect (6). However, results from studies that apply an individualized approach equivalent to Linden et al. (40) to negative affect are still pending at this point.
Saliency of Negative Affective Experiences
Taking into account the meta-analytic finding that depression is associated with altered activation in the saliency network in response to negative affective content (68), Hamilton et al. (42) demonstrated that neurofeedback from subject-specific ROIs in the saliency network can reduce reactivity in the ROIs to negative images. These training effects were additionally reflected in decreased ratings of negative affect in response to the images.
While this study focused on reducing the salience response to negative affective images in a non-neurofeedback transfer task, it did not provide evidence for increased self-regulation performance through neurofeedback. Notably, an early study by Caria et al. (69) showed that healthy participants were able to upregulate aINS activation (a key hub of the salience network), which correlated with increased negative emotion ratings to subsequently presented negative images. Recently, a study by Herwig et al. (70) showed that amygdala reactivity to negative affective images can be downregulated. While these studies suggest that neurofeedback can indeed modulate negative salience responses, future clinical trials are necessary in order to determine whether patients with depression can gain reliable control over their hyperactive salience response through neurofeedback.
Related posts:
Chapter 1 – Brain Imaging Methods in Mood Disorders
Chapter 2 – Neuroanatomical Findings in Unipolar Depression and the Role of the Hippocampus
Chapter 7 – Functional Connectome in Bipolar Disorder
Chapter 8 – Magnetic Resonance Spectroscopy Investigations of Bioenergy and Mitochondrial Function in Mood Disorders
Chapter 15 – Magnetoencephalography Studies in Mood Disorders
Chapter 16 – An Overview of Machine Learning Applications in Mood Disorders
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
