Final Report Summary - BRAINTRAIN (Taking imaging into the therapeutic domain: Self-regulation of brain systems for mental disorders)
Executive Summary:
The BRAINTRAIN project aimed at improving and adapting the methods of real-time fMRI neurofeedback (fMRI-NF) for clinical use, including the combination with electroencephalography (EEG) and the development of standardised procedures for the mapping of brain networks that can be targeted with neurofeedback. Its core component was the exploration of the feasibility and efficacy of fMRI-NF in selected mental and neurodevelopmental disorders that involve motivational, emotional and social neural systems. The ultimate goals of BRAINTRAIN were therefore to:
• Develop new or optimize existing imaging technologies,
• Validate their application as a therapeutic tool to mental and behavioural disorders by integrating imaging data with complementarity knowledge resulting bioinformatics and clinical data,
• Better measure disease progression.
• Develop transfer technologies for fMRI-NF through EEG and serious games.
BRAINTRAIN was innovative in the development of new real-time imaging technologies e.g. new sequences, image reconstruction methods and data analysis software. It also conducted the first formalized clinical testing programme of fMRI-NF in a set of disorders with large socioeconomic and public health impact.
The project started in November 2013 and lasted for five years. The trials have shown good feasibility of fMRI-NF in alcohol dependence, autism, post-traumatic stress disorder, obesity and childhood anxiety.
Project Context and Objectives:
BRAINTRAIN context
Advances in neuroimaging have led to a better knowledge of both mental dysfunction and potential compensatory mechanisms in patients. Major nodes of disordered neural networks are in deep regions of the brain, which makes them difficult to access by electroencephalography or transcranial stimulation. Neuroimaging techniques are therefore essential for the development of non-invasive neuromodulation techniques for mental and behavioural disorders.
Real-time functional magnetic resonance imaging (fMRI) uses magnetic resonance imaging to measure brain activity, by detecting associated changes in blood flow which increases with neuronal activation. fMRI can be used for on-line-monitoring of brain function as well as for selfmodulation of neural processes via interactive training. With the neurofeedback (NF) procedure, patients learn control over brain activity using real-time signals from their own brain.
BRAINTRAIN is based on this idea that real-time functional neuroimaging can be used to train patients to regulate their own brain activity via NF training and thus modulate the brain networks of mental disorder, restore function, improve symptoms and promote resilience. The BRAINTRAIN project brought together the core groups that have been instrumental in the development of methods for real-time functional imaging and fMRI-based NF and have led the initial clinical applications in neuropsychiatric disorders.
BRAINTRAIN objectives
BRAINTRAIN planned to improve and adapt the methods of real-time fMRI NF for clinical use, including the combination with EEG and the development of standardised procedures for the mapping of brain networks that can be targeted with NF. Its core component was be the exploration of the efficacy of fMRI-NF in selected mental and neurodevelopmental disorders that involve motivational, emotional and social neural systems targetable with this technique and pose major public health problems because of their prevalence and hitherto limited treatment options. They are characterised by dysfunctional motivational drives (particularly addiction and binge eating disorder), social communication abilities (particularly autism-spectrum disorder) and emotional regulation (particularly anxiety disorders, including post-traumatic stress disorder), which all have well-established correlates in functional imaging. The development and evaluation of fMRI-NF protocols for these disorders was thus particularly promising on both neurobiological and clinical grounds.
BRAINTRAIN had three main components:
• the development and refinement of methods for the real-time analysis and feedback of fMRI data and combination with other imaging modalities (WP2)
• the adaptation of fMRI mapping techniques to localise disease-relevant networks and development of protocols for their self-regulation through NF (WP3)
• the assessment of feasibility and clinical effects in several mental disorders that are characterised by dysfunctional brain systems for motivation, emotion regulation and social communication and by important therapeutic gaps (autism spectrum disorders, alcohol addiction, post-traumatic stress disorder, childhood anxiety disorders, obesity) (WP4).
BRAINTRAIN partners also wanted to explore the potential transfer of (laboratory-based) imaging feedback training into everyday settings through ambulatory and assistive technologies such as electroencephalography (EEG) and gaming (WP5). The BRAINTRAIN consortium planned to engage with potential users of these technologies (healthcare professionals and providers, medical instrument and software manufacturers, patient and carer associations) through workshops, liaise with regulatory authorities and disseminate findings to the academic and user communities in WP6.
More specifically, BRAINTRAIN objectives were as follows:
WP1 aimed at establishing a Consortium that collaborated efficiently and worked toward the stated objectives, while fulfilling requirements and rules of both the EC grant and consortium agreements.
In order to ensure a smooth functioning of the BRAINTRAIN Consortium, the coordinating institution planned to:
• Establish and maintain the effective management and governance framework for the Consortium, including articulation with an Independent Advisory Board
• Ensure the work and tasks were performed at the highest quality and ethical standards, in a timely manner and within the allotted budget
• Establish a strategy for early identification of risks and threats to the project, thus permitting the implementation of back-up plans for accomplishing stated goals
• Ensure that all actions were performed correctly and within the rules and regulations established by the European Commission and set forth by the Consortium agreement (this included financial and legal management, dissemination rules, distribution and accounting of funds received for the project)
• Inform all partners about the project status, upcoming and emerging issues – this included information about the work plan (adjustments) and all aspects that were important for achieving maximal synergy within the Consortium.
In terms of methods development, the objectives of WP2 were to optimize and consider the entire imaging chain from data acquisition over image reconstruction to data transfer, data analysis and feedback presentation, since the weakest link in this chain would determine the effectiveness of neurofeedback. Partners planned to develop and implement standardised and transferable methods for acquisition, data transfer and analysis of high-quality real-time fMRI data and for transfer to ambulatory devices (EEG).
The specific objectives of this WP2 were set out as follows:
• To optimise and standardise data access and transfer procedures
• To optimise data preprocessing for neurofeedback
• To optimise real-time data analysis software and its link with stimulus presentation/ feedback software
• To provide multivariate pattern and functional connectivity measures for neurofeedback
• To develop tools for direct neurofeedback
• To develop software for cross-modality (fMRI and EEG) neurofeedback
WP3 focused on functional Network mapping and development of specific NF probes. Its objectives were to characterize activity and connectivity indices of functional brain networks involved in motivation, emotion regulation and social communication and to identify the neural probes that can be used for self-modulation of these networks via real-time feedback.
To that end, the specific objectives of WP3 were:
• To develop gaming protocols for fMRI using computerized platforms that effectively recruit major motivation networks mediating reward sensitivity, punishment sensitivity and agency.
• To develop emotion inducing protocols for fMRI that can capture the dynamics of emotional regulation networks involved in reappraisal, suppression or attention bias.
• To develop interactive protocols for fMRI that involved interpersonal communication and relatedness to social cues mediated by empathy related networks such as theory of mind and embodied simulation.
• To identify fMRI probes (activity and/or connectivity) from each network for guiding NF training.
WP4 planned to explore the clinical efficacy of fMRI-based neurofeedback in disorders where dysfunctional activity in motivational networks has been implicated and where current treatments were not satisfactory. The specific objectives of WP4 were to explore feasibility and effects of fMRI-NF in alcohol dependence, obesity, PTSD and ASD and feasibility in adolescent anxiety.
This WP was at the centre of the proposal because it aimed at validating the technology developed in WP2 and WP3 for clinical use. All patient studies of WP4 were Phase 1-2 studies in the framework for the development of complex interventions. Because NF is a complex intervention they were preceded by pilot studies to determine the optimal mode of delivery (e.g. duration of sessions, associated homework protocols), where patient feedback would be incorporated. Some of the trials tested fMRI-NF against an active or placebo control, whereas others were single-group study. The general duration of the intervention was planned to be between one and five months accordingly to clinical rationale.
The general objective of WP5 was to combine fMRI-based neurofeedback with other techniques that would allow for a feedback-guided transfer of the learnt skills to everyday training in ambulatory settings (non-intrusive virtual reality coupled or not with telemetric EEG). BRAINTRAIN partners proposed to use an innovative neurofeedback based approach, based both on their cognitive and methodological research to boost adaptive social decision making. The putative advantages of a virtual reality environment to do the first step of this rehabilitation were grounded on the fact that autistic kids prefer computer interactions to social interactions VR elements were incorporated in the brain-computer-interfaces used by Coimbra and Tel Aviv University for their clinical trials (WP4) but also subjected to further experimental study. WP5 also included the validation of EEG surrogate measures of fMRI-NF (Tel Aviv) and the development of a serious game targeting social skills in autism (Coimbra/ Thales/ UTwente).
WP6 planned a well targeted dissemination effort to maximize impact of the project and release the full exploitation potential of its results. The specific objectives of WP6 were to:
• Develop efficient communication tools for BRAINTRAIN project and reinforce its visibility by creating a graphical chart, a logo and a public website.
• Make sure that all participants would be aware of and understand the Intellectual Property aspects raised by the projects via common rules and a monitoring system.
• Adopt a corporate dissemination strategy, with consortium rules aimed at protecting the IP potential and generated foreground.
• Reach in an early phase of the project a consensus on the exploitation strategy that would satisfy all participants involved in the BRAINTRAIN strategy, avoiding any conflict between the partners.
• Establish as early as possible a network of future end-users, regulatory bodies and the community for 1) reaching out to users to make the application framework adoption as simple as possible and make the final environment match user needs, thus optimising our dissemination effort, 2) preventing bottlenecks, 3) collecting useful feedback from the community, 4) training users and administrators to facilitate their uptake of the project products and services.
• Through industrial partners, to investigate, develop, and if possible, deploy systems and services based on BRAINTRAIN in specific commercial contexts.
• Disseminate the results of BRAINTRAIN project to key stakeholders, including industry, patients associations and carer groups.
Project Results:
WP1 is dedicated to the project coordination and management. This WP thus supported the implementation of scientific and technological activities in the project, contract and finance management for the project duration, internal communication, periodic reporting and organisation of 24 project meetings. This WP did not aim at delivering S&T results.
WP2 led to the following results:
Over the course of the project, real-time data transfer for all MR vendors and multivariate tools for real-time classification of neurofeedback (NF) data were implemented. Tools for computation of real-time functional connectivity have been implemented. Recommendations for 3T MRI scanners for real-time fMRI NF were researched, developed and disseminated. Also, a BOLD sensitivity optimization method has developed, and the development of online EEG analysis tools has been completed. As pilot test, a real-time quality assurance measurement for fMRI was implemented with the opensource image reconstruction framework Gadgetron.
In addition, BrainInnovation released Turbo Brainvoyager 4.0 (TBV 4.0) after extensive beta testing.
TBV 4.0 contains the following new or modified features:
• Easy to use graphical user interface (GUI) with the possibility to visualize the data and statistical results (maps) in multi-slice and single slice view. The visualization on 3D data sets has been extended from AC-PC and Talairach space to MNI space based on user request. Visualization on surface rendered views of the cortical sheet have been improved.
• Import of raw DICOM files has been improved to cover more scanner platforms.
• Interactive GUI during real-time processing allowing to explore the incoming data while running the actual measurement; during real-time processing it is, for example, possible to zoom into single slices, to browse with a 3D cursor in the 3D view, to rotate and zoom 3D models and to select Regions-Of-Interest (ROIs) to display time courses and event-related averaging plots. ROIs can be saved in Talairach and MNI space to be usable across scanning sessions.
• Improved alignment of functional and anatomical data using gradient-based and state-of-the-art boundary-based registration.
• Improved routines for incremental statistical data analysis (recursive least-squares GLM) and incremental event-related averaging.
• Automatic creation of design matrices including confound predictors capturing low frequency drifts. Confound predictors modelling physical and physiological noise can now be added based on white matter or CSF regions of interest.
• Functional connectivity and multi-voxel pattern analysis have been developed for neurofeedback and brain-computer-interface applications.
• Fast incremental 3D motion correction, spatial Gaussian smoothing and drift removal using the design matrix. If GPU available, 3D motion correction may now also use high-quality sinc interpolation to resample functional volumes.
• Statistical threshold and cluster size for overlaid contrast maps can be changed any time during real-time analysis. Regions can now be selected using automatized thresholding procedure.
• Statistical information from multiple overlaid contrast maps can be optionally visualized with separate colors; this "colors code contrasts" mode also highlights voxels with significant conjunctions (intersections) with special colors.
• Storage of fMRI raw data on local hard drive in BrainVoyager format after functional run has been completed allowing an easy transition to in-depth data analysis with offline software such as BrainVoyager or other fMRI software packages (conversion to NIfTI files now supported).
BI also released the Turbo-Satori 1.0 software for real-time Near-Infrared Spectroscopy (NIRS) analysis and finalized the real-time EEG/ fMRI software. For the harmonisation of fMRI data analysis we set up meta-analysis pipelines, which are being utilised by projects from WP4.
MPG has made the SPM-Toolbox for optimizing EPI protocols available on MPG-Github first for BRAINTRAIN members only (later public). MR pulse sequences are available on request if requirements are fulfilled (Siemens collaboration contract, MPG transfer agreement). Gadgetron is publicly available (https://github.com/gadgetron). There is a special image reconstruction in Gadgetron (internal release on Github). We are collecting information about scanners, hard- and software, etc. from partners to prepare for local installations. The direct neurofeedback prototype has been tested.
Additional Activities of MPG during the period:
- Refining MRI Pulse Sequences and Image Reconstruction:
- Ultra-fast fMRI: Assessment of SMS (simultaneous multi-slice) (Todd et al., Frontiers in Neuroscience 2017)
- Ultra-fast fMRI: compressed sensing
WP3 led to the following results:
In Wp3, data on the mapping of emotional networks and different motivational processes has been acquired and analysed. Studies were performed for mapping the functional networks related to cue reactivity reward sensitivity and social cognition, aiming to provide domain specific readouts for various NF training. To improve the accessibility of readout paradigms, an online domain specific paradigm database was created on four domains: Stress, Empathy, Motivation and Emotion Regulation. These paradigms could help identifying the relevant neural probes as well as to employ informative readout measurements. TAU team specifically investigated fMRI probes related to the “social cognition” domain and brain regions with potential applications for rehabilitation in autism-spectrum or social anxiety disorder were identified. Various fMRI pilot studies have been performed by UoC and TAU groups aimed to test the ability to modulate and assess these areas in the context of NF training. In addition, fMRI-based emotion regulation localiser and fMRI-based NF training tasks were developed to assess the feasibility of fMRI-NF of the relevant networks in adolescents. A pilot study performed by Dr Cohen-Kadosh demonstrated the feasibility of such approach in adolescent.
WP4 studied the feasibility of neurofeedback interventions in psychiatric populations. It led to the following results:
BRAINTRAIN was the first programme to study the overall feasibility of delivering neurofeedback interventions in psychiatric populations. All studies met pre-defined success criteria for recruitment, retention and intervention uptake although some variability was evident. We found that this variability relates to the particularly vulnerable nature of some study populations, and the lessons learnt from the BRAINTRAIN studies, which will be published in a paper analysing feasibility metrics across all trials, will inform recruitment and retention strategies of future clinical neurofeedback trials. For instance in the anxiety in adolescents study the high drop- out rate was a result of initial interest from parents which did not translate into scheduling appointments due to conflicting family commitments. This rate, in the healthy cohort, may not reflect the potential willingness to participate in a future clinical sample where both participants and parents are more committed. Additional contacts with participants between visits may also improve retention and uptake in future trials. Eligibility of the population approached was lower than anticipated for the PTSD prevention study: this relates in part to difficulty in identifying relevant populations prior to diagnosis. However, further piloting of identification approaches in a future definitive trial may be worth exploring. In summary the feasibility information, collated systematically for the first time in the area of neurofeedback, points to good feasibility in classical clinical settings, with more challenging recruitment and retention in preventative settings (such as school-based or emergency room-based recruitment).
Neurofeedback mechanisms and intervention methods
The BRAINTRAIN trials also represented the first coordinated programme of neurofeedback research across different neural systems of emotion and motivation (WP3) and clinical disorders (WP4). In terms of technology, we approached the mechanisms of neurofeedback from different angles. The Cardiff, Oxford, Coimbra (neurofeedback) and Tuebingen trials all used fMRI-neurofeedback as primary neuromodulation technique, whereas the TAU trials used fMRI-informed EEG (EEG fingerprinting: EFP) neurofeedback and the Coimbra (BCI) trial used a P300-based brain computer interface (BCI). Because the TAU lab also used pre-/post fMRI and because the neural basis of the P300 is well-understood these trials, too, can be interpreted in terms of fMRI activation effects. Thus, we will be able to explore the neuroplastic effects of neurofeedback interventions across all trials. For the majority of trials which used online fMRI-NF we will also be able to analyse the common areas supporting successful self-regulation (independent of directionality – up-/downregulation of activation levels or functional connectivity) across different paradigms, adding substantially to available meta-analytic evidence.
One unique feature of BRAINTRAIN was the standardisation of outcome measures (where applicable) across trials, which mainly applied to measures of anxiety and depression, but also of cognitive/ psychometric markers that could be predictive of neurofeedback success. This approach will enable meta-analysis of clinical outcome and predictive measures across trials. This document also shows that it is possible to describe and evaluate the feasibility of neurofeedback interventions in terms of state-of-the-art trials research and use common terms to document and evaluate the implementation across clinical areas. This evidence adds to the mapping of current clinical practice that members of the consortium did in collaboration with the international clinical neurofeedback community last year (Randell et al., European Psychiatry 2018).
Efficacy
The BRAINTRAIN trials also provided estimates of efficacy for the primary (disease-specific) outcome measures as well as the secondary outcome measures, some of which were trans-diagnostic (anxiety and depression scales). This design feature of the BRAINTRAIN trails – acquiring common outcome measures across trials/ disease areas – was also a novel and unique element of the consortium’s approach to clinical research.
A) Preliminary evidence on efficacy of Neurofeedback in alcohol dependence
The intervention had a beneficial effect on depression (lowering symptom scores) at month-4 compared to treatment as usual while the effect on anxiety was also positive but smaller. For the alcohol indices measured, both treatment modalities demonstrated beneficial effects. There was a large increase in the number of days abstinent, with the neurofeedback group showing marginal improvements compared to treatment-as-usual. The number of drinks per day and the percentage of days heavy drinking were reduced at follow-up in both groups.
B) Efficacy of Neurofeedback in Autism Spectrum Disorder
For the fMRI study, improvements were observed post-intervention for depression, anxiety and mood score core outcomes. Small improvements were seen for study specific outcomes: an increase in the recognition of expression of emotion (namely in "Fear") in FEEST (the facial emotion recognition test), Sensory/Cognitive Awareness and Health/Physical/Behaviour in ATEC (Autism Treatment Evaluation Checklist) and Communication, Daily Living Skills and total VABS (Vineland Adapative Behavior Scale), indicating a slightly positive effect post-intervention. These differences were maintained at the second follow-up time point.
The results of the BCI study showed similar improvements to the fMRI study for the core outcomes while larger differences were observed for the clinical outcomes. ATEC autism symptoms scores were lower post-intervention while the VABS function scores were higher. These changes indicate a positive effect of the intervention. However the attention scores measured by JAAT did not show improvements neither after intervention nor at the second follow-up time point.
c) Efficacy of Neurofeedback in adolescents with anxiety
The results for the single arm proof of concept study in adolescents indicated that the core outcomes for thought control, emotion regulation, moods and feelings were slightly increased post-intervention compared to baseline. This indicates a slight negative effect of the intervention on mood, however thought control and anxiety scores were improved post-intervention demonstrating a benefit for these outcomes.
d) Efficacy of Neurofeedback in obesity
For the trial of neurofeedback in obesity, participants of both groups succeeded in up-regulating activity of the targeted brain area. However, participants of the control group also showed increased left dlPFC activity during up-regulation. The intervention demonstrated beneficial effects on core outcomes for emotional regulation and mood. However for the depression and anxiety outcomes the intervention showed negative affects compared to control. Study specific scores of eating behaviour were increased in the intervention group compared to control group, indicating a negative effect of the intention on these measures. At follow-up compared to baseline, both groups rated pictures of high-, but not low-calorie foods as less palatable and chose them less frequently. These findings demonstrate that one session of fMRI neurofeedback training enables individuals with increased body weight to up-regulate activity of the left dlPFC.
e) Efficacy of Neurofeedback in PTSD treatment and prevention
In the treatment trial of neurofeedback in PTSD when the two treatment modalities were combined (neutral and trauma EFP-NF) anxiety, PTSD symptom scores were reduced for the intervention group compared to treatment as usual. The scores for depression were marginally higher in the intervention group at follow-up while the score for emotional regulation were mixed with some measures demonstrating slight benefit and other demonstrating a slight negative effect. When comparing neurofeedback in a trauma context with that in a neutral context symptoms score were improved to a greater extent in a trauma context.
In the prevention study PTSD symptom scores were reduced in both groups but the most positive effects were demonstrated in the control (theta-alpha-NF) group. Similar more positive effects in the control group were also observed for the core outcomes for depression and anxiety and the study specific mood regulation outcomes.
Summary
BRAINTRAIN constituted the first ever international effort at systematic investigation of fMRI-neurofeedback in terms of feasibility, brain mechanisms and efficacy across psychiatric disease areas. Feasibility measures were very encouraging regarding the possibility of further clinical trials and ultimately implementation in clinical services. Although preliminary, the efficacy measures also generally pointed in the direction of specific or non-specific (e.g. depression scores) improvement, and would therefore support further exploration of neurofeedback in phase III clinical trials.
WP5 led to the following results:
In WP5 BRAINTRAIN partners used and provided validation for non-intrusive virtual reality displays, and developed fMRI – EEG paradigms to implement transfer technologies as new tools for rehabilitation of autism and PTSD. For autism UoC also developed tools for training goal oriented behavior and joint attention mechanisms to social vs non-social target objects in autism. They also tested EEG and VR technologies for training of social attention and were able to develop gaming technologies coupled with psychophysiological techniques (galvanic skin conductance and eye-tracking). In sum the work first identified brain networks supporting the interpretation of biological facial expressions (with a focus on the posterior Superior Temporal Sulcus) and based on this research, an NF paradigm was designed and optimised in terms of feedback strategies. For PTSD TAU developed both 3D non-specific distressing interface as well as a personalized disease specific feedback interface. This interface was used as an initial phase of the training to enable patients to learn how to down regulate their brain signal under stress but without the threat of re-experiencing their trauma. The personal feedback was composed of the gist of the trauma script obtained from interviewing each patient as part of the diagnostic procedure. The script was recorded by another person (in a second person style) and introduced in the training after approved learning of amygdala down regulation. Successful down regulation of the neural probe resulted in reduced loudness of the audio script. Both groups also applied a new approach for developing fMRI inspired EEG models for NF training. For that different forms of analysis of the EEG signal acquired during fMRI experiments were used to identify correlations between both signals, providing relevant information concerning fMRI to EEG transfer. This shows the potential feasibility of EEG-based neurofeedback for well localized areas even in deep regions such as the amygdala. UoC further developed an EEG-based BCI coupled with VR was developed to train goal oriented behaviour to social vs. non-social target objects in autism. An adapted version of a social cueing training system was validated in a clinical trial. Several virtual reality paradigms were developed to study the functional nature of social impairments in ASD, leading to the definition and prototypical development of several serious games, including one for job interview simulation, for which a market study was implemented.
WP6 led to the following results:
In WP6, BRAINTRAIN partners established a coordinated communication and dissemination strategy. The consortium developed a dedicated website for the project, regularly updated with dissemination activities, publications, project forthcoming workshops, job vacancies. The website was a quite successful tool for dissemination: over the last two years of the project, the number of clicks per article (news) posted ranged from 200 to more than 700, and BRAINTRAIN newsletters, posted on the website, were consulted each up to 800 times. Registration to BRAINTRAIN workshops were embedded in the website, which led to more traffic to the site.
To manage IP issues and check exploitation opportunities, a Business Plan Committee was established and held 10 meetings over the course of the project. The BPC review the results of the project, discussed and updated the Exploitation Plan. The BPC also reviewed potential funding opportunities for the follow-up of the project. Members of the BPC were involved in exchanges with third parties (the UTILE project) to assess the potential project results for further valorisation and commercial exploitation.
Partners worked on the development of an IP and business strategy. To support the exploitation route of the serious game (see task 5c), partners commissioned a market study. It helped partners in assessing market access strategy and prepare the next steps for exploiting the serious game. BRAINTRAIN partners explore future opportunities to maintain networking activities at the consortium level.
In their dissemination strategy, partners paid a particularly attention to feedback from user groups. The investigators collected patient and user group feedback at all stages of the project. Already at the design stage they liaised with service user groups, for example the alcohol services user group at Cwm Taf University Health Board, Wales and autism support groups in Portugal. They also liaised with service user groups and charities during the recruitment stage of the trials, for example through several presentations and discussions at addiction charities in South Wales (e.g. Living Room Project, Recovery Cymru). The feedback from these meetings helped with the fine-tuning of recruitment strategies.
39 scientific publications resulted from BRAINTRAIN during the course of the project.
Potential Impact:
BRAINTRAIN has provided a great amount of evidence for safety, acceptability and feasibility of fMRI-neurofeedback, paving the way for its adoption in clinical services. In WP4, the BRAINTRAIN consortium has developed and evaluated protocols for alcohol dependence, autism, post-traumatic stress disorder, eating disorder and childhood anxiety. In all of these areas there is a huge currently unmet clinical need, and thus new interventions targeted at self-regulation of brain activity could potentially be marketed as therapies to complement existing psychosocial and pharmacological treatment programmes. For example, our 6-session fMRI-neurofeedback training for alcohol could be incorporated into in- or outpatient treatment programmes of addiction services for patients with alcohol dependence in early remission.
Stakeholders for the further development and implementation of these protocols would be clinical services for alcohol dependence, PTSD, autism, eating disorders (obesity) and child and adolescent mental health services (both clinicians and managers/ commissioners of health services). All these clinical areas address huge socioeconomic problems. For example, harmful use of alcohol has been estimated to lead to economic costs in the range between 1.3 and 3.3% of GDP (https://ec.europa.eu/health/sites/health/files/social_determinants/docs/hepp_screport_alcohol_en.pdf) and thus around 400 billion Euro annually in the EU. Even interventions with small benefit, for example a small reduction in the rate of relapse after alcohol detoxification, could have very large clinical and socioeconomic benefits. Further innovation in treatment research is therefore needed, and our survey of the international community (Randell et al., 2018) and the spectrum of abstract submissions to the 3rd international real-time fMRI and neurofeedback conference in Nara, Japan (Nov. 2017) indicate that the international interest in clinical applications of fMRI-NF is growing, and several new trials are being set up. Through published protocols and advice provided to other investigators the BRAINTRAIN consortium has established higher standards for methodological rigour and standardisation which are now being adopted across the international fMRI-neurofeedback community. We would therefore expect that the next wave of larger efficacy studies in the next 3-5 years will provide the evidence to support wider clinical use of fMRI-neurofeedback for at least two psychiatric indications. In the meantime we also envisage a wider uptake of fMRI-neurofeedback because of its good safety and acceptability profile in experimental treatment settings, which will benefit otherwise treatment-refractory patients.
WP2 constituted the first effort worldwide to optimise the full pipeline of real-time imaging for clinical neurofeedback (data acquisition, data transfer, data processing, feedback interface, standardised analyses). This has resulted in tangible outputs in terms of non-commercial protocol sharing by MPG and commercial developments (Turbo Brainvoyager MED, Turbo-Satori) by BI. TBV is the most widely used software package for real-time fMRI/ neurofeedback worldwide.
The network mapping paradigms developed in WP3 (disseminated through open accessible published output) are already being used by other groups to inform their clinical neurofeedback protocol design (e.g. in a trial on eating disorders at the University of Leipzig, Germany).
With regards to WP5, because there are still no available therapies to treat the core symptoms of autism (only medications are available for secondary features such as anxiety or aggressive behaviour) and neuromodulation approaches based on neurostimulation have also been disappointing so far our approach concerning neurofeedback, EEG-based BCI and serious games is both novel and promising for the development of new integrated treatments, which can be taken up by neuroscience centres across the world.
At last, BRAINTRAIN led a series of versatile actions to disseminate the project results. Partners published 39 peered-review papers, and have several publications under review, and presented BRAINTRAIN research and results in more than 130 dissemination events, including conferences, symposium, workshops, patient association forum. Several publications and presentations are planned after the project end. To maximize impact of the project, BRAINTRAIN website will also be active for another two years after the project ends.
During the course of the project BRAINTRAIN also held successful workshops targeting specific users and stakeholders. The project partners organized six workshop and dissemination events, in various countries (Israel, Germany, UK, Portugal, the Netherlands), reaching out to more than 400 people.They included 1. A standardisation workshop with industry, to get in close contact with MRI hardware vendors, and establish links with industry for further commercialisation options. 2. A training workshop for researchers and clinicians, to address the expectations of patients and service providers. 3. A liaison workshop with industry, focusing on transfer technologies such as EEG and brain stimulation. 4. A liaison event for patient/carer groups and associations, to raise awareness of clinical potential of NF and prepare further large-scale clinical trials, 5. A general workshop to address advances in neurofeedback research and practices. 6. A training workshop on regulatory issues.
At last, BRAINTRAIN partners engaged a broader audience in their activities, targeting general public during National Autism Day Commemorations (Portugal, 2016), policy makers (Café de Ciencia in Portuguse Parliament, 2014; Poster presentation to the Welsh Government Funders events), and reaching out a large audience in TV clips (BBC Wales TV programme “Do I Drink Too Much” on alcohol dependence, 2016, Brain Imaging clip on Porto Canal, 2018).
Exploitation routes for the project are diverse. The BRAINTRAIN consortium has already produced several technological and research deliverables that are promising targets for exploitation. These can broadly be divided into the areas of advanced MR imaging technologies, transfer technologies, clinical trial standards and improvement of service structures. All of these areas are subject to different practices, rules and regulations, and the consortium is building up expertise (for example, through stakeholder consultation and workshops) in these different exploitation scenarios.
Various exploitations routes have been planned: market studies, contact with companies, direct commercial exploitation, and non-commercial exploitation (harmonization of neurofeedback protocols; software sharing).
List of Websites:
http://www.braintrainproject.eu/
The BRAINTRAIN project aimed at improving and adapting the methods of real-time fMRI neurofeedback (fMRI-NF) for clinical use, including the combination with electroencephalography (EEG) and the development of standardised procedures for the mapping of brain networks that can be targeted with neurofeedback. Its core component was the exploration of the feasibility and efficacy of fMRI-NF in selected mental and neurodevelopmental disorders that involve motivational, emotional and social neural systems. The ultimate goals of BRAINTRAIN were therefore to:
• Develop new or optimize existing imaging technologies,
• Validate their application as a therapeutic tool to mental and behavioural disorders by integrating imaging data with complementarity knowledge resulting bioinformatics and clinical data,
• Better measure disease progression.
• Develop transfer technologies for fMRI-NF through EEG and serious games.
BRAINTRAIN was innovative in the development of new real-time imaging technologies e.g. new sequences, image reconstruction methods and data analysis software. It also conducted the first formalized clinical testing programme of fMRI-NF in a set of disorders with large socioeconomic and public health impact.
The project started in November 2013 and lasted for five years. The trials have shown good feasibility of fMRI-NF in alcohol dependence, autism, post-traumatic stress disorder, obesity and childhood anxiety.
Project Context and Objectives:
BRAINTRAIN context
Advances in neuroimaging have led to a better knowledge of both mental dysfunction and potential compensatory mechanisms in patients. Major nodes of disordered neural networks are in deep regions of the brain, which makes them difficult to access by electroencephalography or transcranial stimulation. Neuroimaging techniques are therefore essential for the development of non-invasive neuromodulation techniques for mental and behavioural disorders.
Real-time functional magnetic resonance imaging (fMRI) uses magnetic resonance imaging to measure brain activity, by detecting associated changes in blood flow which increases with neuronal activation. fMRI can be used for on-line-monitoring of brain function as well as for selfmodulation of neural processes via interactive training. With the neurofeedback (NF) procedure, patients learn control over brain activity using real-time signals from their own brain.
BRAINTRAIN is based on this idea that real-time functional neuroimaging can be used to train patients to regulate their own brain activity via NF training and thus modulate the brain networks of mental disorder, restore function, improve symptoms and promote resilience. The BRAINTRAIN project brought together the core groups that have been instrumental in the development of methods for real-time functional imaging and fMRI-based NF and have led the initial clinical applications in neuropsychiatric disorders.
BRAINTRAIN objectives
BRAINTRAIN planned to improve and adapt the methods of real-time fMRI NF for clinical use, including the combination with EEG and the development of standardised procedures for the mapping of brain networks that can be targeted with NF. Its core component was be the exploration of the efficacy of fMRI-NF in selected mental and neurodevelopmental disorders that involve motivational, emotional and social neural systems targetable with this technique and pose major public health problems because of their prevalence and hitherto limited treatment options. They are characterised by dysfunctional motivational drives (particularly addiction and binge eating disorder), social communication abilities (particularly autism-spectrum disorder) and emotional regulation (particularly anxiety disorders, including post-traumatic stress disorder), which all have well-established correlates in functional imaging. The development and evaluation of fMRI-NF protocols for these disorders was thus particularly promising on both neurobiological and clinical grounds.
BRAINTRAIN had three main components:
• the development and refinement of methods for the real-time analysis and feedback of fMRI data and combination with other imaging modalities (WP2)
• the adaptation of fMRI mapping techniques to localise disease-relevant networks and development of protocols for their self-regulation through NF (WP3)
• the assessment of feasibility and clinical effects in several mental disorders that are characterised by dysfunctional brain systems for motivation, emotion regulation and social communication and by important therapeutic gaps (autism spectrum disorders, alcohol addiction, post-traumatic stress disorder, childhood anxiety disorders, obesity) (WP4).
BRAINTRAIN partners also wanted to explore the potential transfer of (laboratory-based) imaging feedback training into everyday settings through ambulatory and assistive technologies such as electroencephalography (EEG) and gaming (WP5). The BRAINTRAIN consortium planned to engage with potential users of these technologies (healthcare professionals and providers, medical instrument and software manufacturers, patient and carer associations) through workshops, liaise with regulatory authorities and disseminate findings to the academic and user communities in WP6.
More specifically, BRAINTRAIN objectives were as follows:
WP1 aimed at establishing a Consortium that collaborated efficiently and worked toward the stated objectives, while fulfilling requirements and rules of both the EC grant and consortium agreements.
In order to ensure a smooth functioning of the BRAINTRAIN Consortium, the coordinating institution planned to:
• Establish and maintain the effective management and governance framework for the Consortium, including articulation with an Independent Advisory Board
• Ensure the work and tasks were performed at the highest quality and ethical standards, in a timely manner and within the allotted budget
• Establish a strategy for early identification of risks and threats to the project, thus permitting the implementation of back-up plans for accomplishing stated goals
• Ensure that all actions were performed correctly and within the rules and regulations established by the European Commission and set forth by the Consortium agreement (this included financial and legal management, dissemination rules, distribution and accounting of funds received for the project)
• Inform all partners about the project status, upcoming and emerging issues – this included information about the work plan (adjustments) and all aspects that were important for achieving maximal synergy within the Consortium.
In terms of methods development, the objectives of WP2 were to optimize and consider the entire imaging chain from data acquisition over image reconstruction to data transfer, data analysis and feedback presentation, since the weakest link in this chain would determine the effectiveness of neurofeedback. Partners planned to develop and implement standardised and transferable methods for acquisition, data transfer and analysis of high-quality real-time fMRI data and for transfer to ambulatory devices (EEG).
The specific objectives of this WP2 were set out as follows:
• To optimise and standardise data access and transfer procedures
• To optimise data preprocessing for neurofeedback
• To optimise real-time data analysis software and its link with stimulus presentation/ feedback software
• To provide multivariate pattern and functional connectivity measures for neurofeedback
• To develop tools for direct neurofeedback
• To develop software for cross-modality (fMRI and EEG) neurofeedback
WP3 focused on functional Network mapping and development of specific NF probes. Its objectives were to characterize activity and connectivity indices of functional brain networks involved in motivation, emotion regulation and social communication and to identify the neural probes that can be used for self-modulation of these networks via real-time feedback.
To that end, the specific objectives of WP3 were:
• To develop gaming protocols for fMRI using computerized platforms that effectively recruit major motivation networks mediating reward sensitivity, punishment sensitivity and agency.
• To develop emotion inducing protocols for fMRI that can capture the dynamics of emotional regulation networks involved in reappraisal, suppression or attention bias.
• To develop interactive protocols for fMRI that involved interpersonal communication and relatedness to social cues mediated by empathy related networks such as theory of mind and embodied simulation.
• To identify fMRI probes (activity and/or connectivity) from each network for guiding NF training.
WP4 planned to explore the clinical efficacy of fMRI-based neurofeedback in disorders where dysfunctional activity in motivational networks has been implicated and where current treatments were not satisfactory. The specific objectives of WP4 were to explore feasibility and effects of fMRI-NF in alcohol dependence, obesity, PTSD and ASD and feasibility in adolescent anxiety.
This WP was at the centre of the proposal because it aimed at validating the technology developed in WP2 and WP3 for clinical use. All patient studies of WP4 were Phase 1-2 studies in the framework for the development of complex interventions. Because NF is a complex intervention they were preceded by pilot studies to determine the optimal mode of delivery (e.g. duration of sessions, associated homework protocols), where patient feedback would be incorporated. Some of the trials tested fMRI-NF against an active or placebo control, whereas others were single-group study. The general duration of the intervention was planned to be between one and five months accordingly to clinical rationale.
The general objective of WP5 was to combine fMRI-based neurofeedback with other techniques that would allow for a feedback-guided transfer of the learnt skills to everyday training in ambulatory settings (non-intrusive virtual reality coupled or not with telemetric EEG). BRAINTRAIN partners proposed to use an innovative neurofeedback based approach, based both on their cognitive and methodological research to boost adaptive social decision making. The putative advantages of a virtual reality environment to do the first step of this rehabilitation were grounded on the fact that autistic kids prefer computer interactions to social interactions VR elements were incorporated in the brain-computer-interfaces used by Coimbra and Tel Aviv University for their clinical trials (WP4) but also subjected to further experimental study. WP5 also included the validation of EEG surrogate measures of fMRI-NF (Tel Aviv) and the development of a serious game targeting social skills in autism (Coimbra/ Thales/ UTwente).
WP6 planned a well targeted dissemination effort to maximize impact of the project and release the full exploitation potential of its results. The specific objectives of WP6 were to:
• Develop efficient communication tools for BRAINTRAIN project and reinforce its visibility by creating a graphical chart, a logo and a public website.
• Make sure that all participants would be aware of and understand the Intellectual Property aspects raised by the projects via common rules and a monitoring system.
• Adopt a corporate dissemination strategy, with consortium rules aimed at protecting the IP potential and generated foreground.
• Reach in an early phase of the project a consensus on the exploitation strategy that would satisfy all participants involved in the BRAINTRAIN strategy, avoiding any conflict between the partners.
• Establish as early as possible a network of future end-users, regulatory bodies and the community for 1) reaching out to users to make the application framework adoption as simple as possible and make the final environment match user needs, thus optimising our dissemination effort, 2) preventing bottlenecks, 3) collecting useful feedback from the community, 4) training users and administrators to facilitate their uptake of the project products and services.
• Through industrial partners, to investigate, develop, and if possible, deploy systems and services based on BRAINTRAIN in specific commercial contexts.
• Disseminate the results of BRAINTRAIN project to key stakeholders, including industry, patients associations and carer groups.
Project Results:
WP1 is dedicated to the project coordination and management. This WP thus supported the implementation of scientific and technological activities in the project, contract and finance management for the project duration, internal communication, periodic reporting and organisation of 24 project meetings. This WP did not aim at delivering S&T results.
WP2 led to the following results:
Over the course of the project, real-time data transfer for all MR vendors and multivariate tools for real-time classification of neurofeedback (NF) data were implemented. Tools for computation of real-time functional connectivity have been implemented. Recommendations for 3T MRI scanners for real-time fMRI NF were researched, developed and disseminated. Also, a BOLD sensitivity optimization method has developed, and the development of online EEG analysis tools has been completed. As pilot test, a real-time quality assurance measurement for fMRI was implemented with the opensource image reconstruction framework Gadgetron.
In addition, BrainInnovation released Turbo Brainvoyager 4.0 (TBV 4.0) after extensive beta testing.
TBV 4.0 contains the following new or modified features:
• Easy to use graphical user interface (GUI) with the possibility to visualize the data and statistical results (maps) in multi-slice and single slice view. The visualization on 3D data sets has been extended from AC-PC and Talairach space to MNI space based on user request. Visualization on surface rendered views of the cortical sheet have been improved.
• Import of raw DICOM files has been improved to cover more scanner platforms.
• Interactive GUI during real-time processing allowing to explore the incoming data while running the actual measurement; during real-time processing it is, for example, possible to zoom into single slices, to browse with a 3D cursor in the 3D view, to rotate and zoom 3D models and to select Regions-Of-Interest (ROIs) to display time courses and event-related averaging plots. ROIs can be saved in Talairach and MNI space to be usable across scanning sessions.
• Improved alignment of functional and anatomical data using gradient-based and state-of-the-art boundary-based registration.
• Improved routines for incremental statistical data analysis (recursive least-squares GLM) and incremental event-related averaging.
• Automatic creation of design matrices including confound predictors capturing low frequency drifts. Confound predictors modelling physical and physiological noise can now be added based on white matter or CSF regions of interest.
• Functional connectivity and multi-voxel pattern analysis have been developed for neurofeedback and brain-computer-interface applications.
• Fast incremental 3D motion correction, spatial Gaussian smoothing and drift removal using the design matrix. If GPU available, 3D motion correction may now also use high-quality sinc interpolation to resample functional volumes.
• Statistical threshold and cluster size for overlaid contrast maps can be changed any time during real-time analysis. Regions can now be selected using automatized thresholding procedure.
• Statistical information from multiple overlaid contrast maps can be optionally visualized with separate colors; this "colors code contrasts" mode also highlights voxels with significant conjunctions (intersections) with special colors.
• Storage of fMRI raw data on local hard drive in BrainVoyager format after functional run has been completed allowing an easy transition to in-depth data analysis with offline software such as BrainVoyager or other fMRI software packages (conversion to NIfTI files now supported).
BI also released the Turbo-Satori 1.0 software for real-time Near-Infrared Spectroscopy (NIRS) analysis and finalized the real-time EEG/ fMRI software. For the harmonisation of fMRI data analysis we set up meta-analysis pipelines, which are being utilised by projects from WP4.
MPG has made the SPM-Toolbox for optimizing EPI protocols available on MPG-Github first for BRAINTRAIN members only (later public). MR pulse sequences are available on request if requirements are fulfilled (Siemens collaboration contract, MPG transfer agreement). Gadgetron is publicly available (https://github.com/gadgetron). There is a special image reconstruction in Gadgetron (internal release on Github). We are collecting information about scanners, hard- and software, etc. from partners to prepare for local installations. The direct neurofeedback prototype has been tested.
Additional Activities of MPG during the period:
- Refining MRI Pulse Sequences and Image Reconstruction:
- Ultra-fast fMRI: Assessment of SMS (simultaneous multi-slice) (Todd et al., Frontiers in Neuroscience 2017)
- Ultra-fast fMRI: compressed sensing
WP3 led to the following results:
In Wp3, data on the mapping of emotional networks and different motivational processes has been acquired and analysed. Studies were performed for mapping the functional networks related to cue reactivity reward sensitivity and social cognition, aiming to provide domain specific readouts for various NF training. To improve the accessibility of readout paradigms, an online domain specific paradigm database was created on four domains: Stress, Empathy, Motivation and Emotion Regulation. These paradigms could help identifying the relevant neural probes as well as to employ informative readout measurements. TAU team specifically investigated fMRI probes related to the “social cognition” domain and brain regions with potential applications for rehabilitation in autism-spectrum or social anxiety disorder were identified. Various fMRI pilot studies have been performed by UoC and TAU groups aimed to test the ability to modulate and assess these areas in the context of NF training. In addition, fMRI-based emotion regulation localiser and fMRI-based NF training tasks were developed to assess the feasibility of fMRI-NF of the relevant networks in adolescents. A pilot study performed by Dr Cohen-Kadosh demonstrated the feasibility of such approach in adolescent.
WP4 studied the feasibility of neurofeedback interventions in psychiatric populations. It led to the following results:
BRAINTRAIN was the first programme to study the overall feasibility of delivering neurofeedback interventions in psychiatric populations. All studies met pre-defined success criteria for recruitment, retention and intervention uptake although some variability was evident. We found that this variability relates to the particularly vulnerable nature of some study populations, and the lessons learnt from the BRAINTRAIN studies, which will be published in a paper analysing feasibility metrics across all trials, will inform recruitment and retention strategies of future clinical neurofeedback trials. For instance in the anxiety in adolescents study the high drop- out rate was a result of initial interest from parents which did not translate into scheduling appointments due to conflicting family commitments. This rate, in the healthy cohort, may not reflect the potential willingness to participate in a future clinical sample where both participants and parents are more committed. Additional contacts with participants between visits may also improve retention and uptake in future trials. Eligibility of the population approached was lower than anticipated for the PTSD prevention study: this relates in part to difficulty in identifying relevant populations prior to diagnosis. However, further piloting of identification approaches in a future definitive trial may be worth exploring. In summary the feasibility information, collated systematically for the first time in the area of neurofeedback, points to good feasibility in classical clinical settings, with more challenging recruitment and retention in preventative settings (such as school-based or emergency room-based recruitment).
Neurofeedback mechanisms and intervention methods
The BRAINTRAIN trials also represented the first coordinated programme of neurofeedback research across different neural systems of emotion and motivation (WP3) and clinical disorders (WP4). In terms of technology, we approached the mechanisms of neurofeedback from different angles. The Cardiff, Oxford, Coimbra (neurofeedback) and Tuebingen trials all used fMRI-neurofeedback as primary neuromodulation technique, whereas the TAU trials used fMRI-informed EEG (EEG fingerprinting: EFP) neurofeedback and the Coimbra (BCI) trial used a P300-based brain computer interface (BCI). Because the TAU lab also used pre-/post fMRI and because the neural basis of the P300 is well-understood these trials, too, can be interpreted in terms of fMRI activation effects. Thus, we will be able to explore the neuroplastic effects of neurofeedback interventions across all trials. For the majority of trials which used online fMRI-NF we will also be able to analyse the common areas supporting successful self-regulation (independent of directionality – up-/downregulation of activation levels or functional connectivity) across different paradigms, adding substantially to available meta-analytic evidence.
One unique feature of BRAINTRAIN was the standardisation of outcome measures (where applicable) across trials, which mainly applied to measures of anxiety and depression, but also of cognitive/ psychometric markers that could be predictive of neurofeedback success. This approach will enable meta-analysis of clinical outcome and predictive measures across trials. This document also shows that it is possible to describe and evaluate the feasibility of neurofeedback interventions in terms of state-of-the-art trials research and use common terms to document and evaluate the implementation across clinical areas. This evidence adds to the mapping of current clinical practice that members of the consortium did in collaboration with the international clinical neurofeedback community last year (Randell et al., European Psychiatry 2018).
Efficacy
The BRAINTRAIN trials also provided estimates of efficacy for the primary (disease-specific) outcome measures as well as the secondary outcome measures, some of which were trans-diagnostic (anxiety and depression scales). This design feature of the BRAINTRAIN trails – acquiring common outcome measures across trials/ disease areas – was also a novel and unique element of the consortium’s approach to clinical research.
A) Preliminary evidence on efficacy of Neurofeedback in alcohol dependence
The intervention had a beneficial effect on depression (lowering symptom scores) at month-4 compared to treatment as usual while the effect on anxiety was also positive but smaller. For the alcohol indices measured, both treatment modalities demonstrated beneficial effects. There was a large increase in the number of days abstinent, with the neurofeedback group showing marginal improvements compared to treatment-as-usual. The number of drinks per day and the percentage of days heavy drinking were reduced at follow-up in both groups.
B) Efficacy of Neurofeedback in Autism Spectrum Disorder
For the fMRI study, improvements were observed post-intervention for depression, anxiety and mood score core outcomes. Small improvements were seen for study specific outcomes: an increase in the recognition of expression of emotion (namely in "Fear") in FEEST (the facial emotion recognition test), Sensory/Cognitive Awareness and Health/Physical/Behaviour in ATEC (Autism Treatment Evaluation Checklist) and Communication, Daily Living Skills and total VABS (Vineland Adapative Behavior Scale), indicating a slightly positive effect post-intervention. These differences were maintained at the second follow-up time point.
The results of the BCI study showed similar improvements to the fMRI study for the core outcomes while larger differences were observed for the clinical outcomes. ATEC autism symptoms scores were lower post-intervention while the VABS function scores were higher. These changes indicate a positive effect of the intervention. However the attention scores measured by JAAT did not show improvements neither after intervention nor at the second follow-up time point.
c) Efficacy of Neurofeedback in adolescents with anxiety
The results for the single arm proof of concept study in adolescents indicated that the core outcomes for thought control, emotion regulation, moods and feelings were slightly increased post-intervention compared to baseline. This indicates a slight negative effect of the intervention on mood, however thought control and anxiety scores were improved post-intervention demonstrating a benefit for these outcomes.
d) Efficacy of Neurofeedback in obesity
For the trial of neurofeedback in obesity, participants of both groups succeeded in up-regulating activity of the targeted brain area. However, participants of the control group also showed increased left dlPFC activity during up-regulation. The intervention demonstrated beneficial effects on core outcomes for emotional regulation and mood. However for the depression and anxiety outcomes the intervention showed negative affects compared to control. Study specific scores of eating behaviour were increased in the intervention group compared to control group, indicating a negative effect of the intention on these measures. At follow-up compared to baseline, both groups rated pictures of high-, but not low-calorie foods as less palatable and chose them less frequently. These findings demonstrate that one session of fMRI neurofeedback training enables individuals with increased body weight to up-regulate activity of the left dlPFC.
e) Efficacy of Neurofeedback in PTSD treatment and prevention
In the treatment trial of neurofeedback in PTSD when the two treatment modalities were combined (neutral and trauma EFP-NF) anxiety, PTSD symptom scores were reduced for the intervention group compared to treatment as usual. The scores for depression were marginally higher in the intervention group at follow-up while the score for emotional regulation were mixed with some measures demonstrating slight benefit and other demonstrating a slight negative effect. When comparing neurofeedback in a trauma context with that in a neutral context symptoms score were improved to a greater extent in a trauma context.
In the prevention study PTSD symptom scores were reduced in both groups but the most positive effects were demonstrated in the control (theta-alpha-NF) group. Similar more positive effects in the control group were also observed for the core outcomes for depression and anxiety and the study specific mood regulation outcomes.
Summary
BRAINTRAIN constituted the first ever international effort at systematic investigation of fMRI-neurofeedback in terms of feasibility, brain mechanisms and efficacy across psychiatric disease areas. Feasibility measures were very encouraging regarding the possibility of further clinical trials and ultimately implementation in clinical services. Although preliminary, the efficacy measures also generally pointed in the direction of specific or non-specific (e.g. depression scores) improvement, and would therefore support further exploration of neurofeedback in phase III clinical trials.
WP5 led to the following results:
In WP5 BRAINTRAIN partners used and provided validation for non-intrusive virtual reality displays, and developed fMRI – EEG paradigms to implement transfer technologies as new tools for rehabilitation of autism and PTSD. For autism UoC also developed tools for training goal oriented behavior and joint attention mechanisms to social vs non-social target objects in autism. They also tested EEG and VR technologies for training of social attention and were able to develop gaming technologies coupled with psychophysiological techniques (galvanic skin conductance and eye-tracking). In sum the work first identified brain networks supporting the interpretation of biological facial expressions (with a focus on the posterior Superior Temporal Sulcus) and based on this research, an NF paradigm was designed and optimised in terms of feedback strategies. For PTSD TAU developed both 3D non-specific distressing interface as well as a personalized disease specific feedback interface. This interface was used as an initial phase of the training to enable patients to learn how to down regulate their brain signal under stress but without the threat of re-experiencing their trauma. The personal feedback was composed of the gist of the trauma script obtained from interviewing each patient as part of the diagnostic procedure. The script was recorded by another person (in a second person style) and introduced in the training after approved learning of amygdala down regulation. Successful down regulation of the neural probe resulted in reduced loudness of the audio script. Both groups also applied a new approach for developing fMRI inspired EEG models for NF training. For that different forms of analysis of the EEG signal acquired during fMRI experiments were used to identify correlations between both signals, providing relevant information concerning fMRI to EEG transfer. This shows the potential feasibility of EEG-based neurofeedback for well localized areas even in deep regions such as the amygdala. UoC further developed an EEG-based BCI coupled with VR was developed to train goal oriented behaviour to social vs. non-social target objects in autism. An adapted version of a social cueing training system was validated in a clinical trial. Several virtual reality paradigms were developed to study the functional nature of social impairments in ASD, leading to the definition and prototypical development of several serious games, including one for job interview simulation, for which a market study was implemented.
WP6 led to the following results:
In WP6, BRAINTRAIN partners established a coordinated communication and dissemination strategy. The consortium developed a dedicated website for the project, regularly updated with dissemination activities, publications, project forthcoming workshops, job vacancies. The website was a quite successful tool for dissemination: over the last two years of the project, the number of clicks per article (news) posted ranged from 200 to more than 700, and BRAINTRAIN newsletters, posted on the website, were consulted each up to 800 times. Registration to BRAINTRAIN workshops were embedded in the website, which led to more traffic to the site.
To manage IP issues and check exploitation opportunities, a Business Plan Committee was established and held 10 meetings over the course of the project. The BPC review the results of the project, discussed and updated the Exploitation Plan. The BPC also reviewed potential funding opportunities for the follow-up of the project. Members of the BPC were involved in exchanges with third parties (the UTILE project) to assess the potential project results for further valorisation and commercial exploitation.
Partners worked on the development of an IP and business strategy. To support the exploitation route of the serious game (see task 5c), partners commissioned a market study. It helped partners in assessing market access strategy and prepare the next steps for exploiting the serious game. BRAINTRAIN partners explore future opportunities to maintain networking activities at the consortium level.
In their dissemination strategy, partners paid a particularly attention to feedback from user groups. The investigators collected patient and user group feedback at all stages of the project. Already at the design stage they liaised with service user groups, for example the alcohol services user group at Cwm Taf University Health Board, Wales and autism support groups in Portugal. They also liaised with service user groups and charities during the recruitment stage of the trials, for example through several presentations and discussions at addiction charities in South Wales (e.g. Living Room Project, Recovery Cymru). The feedback from these meetings helped with the fine-tuning of recruitment strategies.
39 scientific publications resulted from BRAINTRAIN during the course of the project.
Potential Impact:
BRAINTRAIN has provided a great amount of evidence for safety, acceptability and feasibility of fMRI-neurofeedback, paving the way for its adoption in clinical services. In WP4, the BRAINTRAIN consortium has developed and evaluated protocols for alcohol dependence, autism, post-traumatic stress disorder, eating disorder and childhood anxiety. In all of these areas there is a huge currently unmet clinical need, and thus new interventions targeted at self-regulation of brain activity could potentially be marketed as therapies to complement existing psychosocial and pharmacological treatment programmes. For example, our 6-session fMRI-neurofeedback training for alcohol could be incorporated into in- or outpatient treatment programmes of addiction services for patients with alcohol dependence in early remission.
Stakeholders for the further development and implementation of these protocols would be clinical services for alcohol dependence, PTSD, autism, eating disorders (obesity) and child and adolescent mental health services (both clinicians and managers/ commissioners of health services). All these clinical areas address huge socioeconomic problems. For example, harmful use of alcohol has been estimated to lead to economic costs in the range between 1.3 and 3.3% of GDP (https://ec.europa.eu/health/sites/health/files/social_determinants/docs/hepp_screport_alcohol_en.pdf) and thus around 400 billion Euro annually in the EU. Even interventions with small benefit, for example a small reduction in the rate of relapse after alcohol detoxification, could have very large clinical and socioeconomic benefits. Further innovation in treatment research is therefore needed, and our survey of the international community (Randell et al., 2018) and the spectrum of abstract submissions to the 3rd international real-time fMRI and neurofeedback conference in Nara, Japan (Nov. 2017) indicate that the international interest in clinical applications of fMRI-NF is growing, and several new trials are being set up. Through published protocols and advice provided to other investigators the BRAINTRAIN consortium has established higher standards for methodological rigour and standardisation which are now being adopted across the international fMRI-neurofeedback community. We would therefore expect that the next wave of larger efficacy studies in the next 3-5 years will provide the evidence to support wider clinical use of fMRI-neurofeedback for at least two psychiatric indications. In the meantime we also envisage a wider uptake of fMRI-neurofeedback because of its good safety and acceptability profile in experimental treatment settings, which will benefit otherwise treatment-refractory patients.
WP2 constituted the first effort worldwide to optimise the full pipeline of real-time imaging for clinical neurofeedback (data acquisition, data transfer, data processing, feedback interface, standardised analyses). This has resulted in tangible outputs in terms of non-commercial protocol sharing by MPG and commercial developments (Turbo Brainvoyager MED, Turbo-Satori) by BI. TBV is the most widely used software package for real-time fMRI/ neurofeedback worldwide.
The network mapping paradigms developed in WP3 (disseminated through open accessible published output) are already being used by other groups to inform their clinical neurofeedback protocol design (e.g. in a trial on eating disorders at the University of Leipzig, Germany).
With regards to WP5, because there are still no available therapies to treat the core symptoms of autism (only medications are available for secondary features such as anxiety or aggressive behaviour) and neuromodulation approaches based on neurostimulation have also been disappointing so far our approach concerning neurofeedback, EEG-based BCI and serious games is both novel and promising for the development of new integrated treatments, which can be taken up by neuroscience centres across the world.
At last, BRAINTRAIN led a series of versatile actions to disseminate the project results. Partners published 39 peered-review papers, and have several publications under review, and presented BRAINTRAIN research and results in more than 130 dissemination events, including conferences, symposium, workshops, patient association forum. Several publications and presentations are planned after the project end. To maximize impact of the project, BRAINTRAIN website will also be active for another two years after the project ends.
During the course of the project BRAINTRAIN also held successful workshops targeting specific users and stakeholders. The project partners organized six workshop and dissemination events, in various countries (Israel, Germany, UK, Portugal, the Netherlands), reaching out to more than 400 people.They included 1. A standardisation workshop with industry, to get in close contact with MRI hardware vendors, and establish links with industry for further commercialisation options. 2. A training workshop for researchers and clinicians, to address the expectations of patients and service providers. 3. A liaison workshop with industry, focusing on transfer technologies such as EEG and brain stimulation. 4. A liaison event for patient/carer groups and associations, to raise awareness of clinical potential of NF and prepare further large-scale clinical trials, 5. A general workshop to address advances in neurofeedback research and practices. 6. A training workshop on regulatory issues.
At last, BRAINTRAIN partners engaged a broader audience in their activities, targeting general public during National Autism Day Commemorations (Portugal, 2016), policy makers (Café de Ciencia in Portuguse Parliament, 2014; Poster presentation to the Welsh Government Funders events), and reaching out a large audience in TV clips (BBC Wales TV programme “Do I Drink Too Much” on alcohol dependence, 2016, Brain Imaging clip on Porto Canal, 2018).
Exploitation routes for the project are diverse. The BRAINTRAIN consortium has already produced several technological and research deliverables that are promising targets for exploitation. These can broadly be divided into the areas of advanced MR imaging technologies, transfer technologies, clinical trial standards and improvement of service structures. All of these areas are subject to different practices, rules and regulations, and the consortium is building up expertise (for example, through stakeholder consultation and workshops) in these different exploitation scenarios.
Various exploitations routes have been planned: market studies, contact with companies, direct commercial exploitation, and non-commercial exploitation (harmonization of neurofeedback protocols; software sharing).
List of Websites:
http://www.braintrainproject.eu/