Predictors of real-time fMRI neurofeedback performance and improvement - A machine learning mega-analysis

Real-time fMRI neurofeedback is an increasingly popular neuroimaging technique that allows an individual to gain control over his/her own brain signals, which can lead to improvements in behavior in healthy participants as well as to improvements of clinical symptoms in patient populations. However, a considerably large ratio of participants undergoing neurofeedback training do not learn to control their own brain signals and, consequently, do not benefit from neurofeedback interventions, which limits clinical efficacy of neurofeedback interventions. As neurofeedback success varies between studies and participants, it is important to identify factors that might influence neurofeedback success. Here, for the first time, we employed a big data machine learning approach to investigate the influence of 20 different design-specific (e.g. activity vs. connectivity feedback), region of interest-specific (e.g. cortical vs. subcortical) and subject-specific factors (e.g. age) on neurofeedback performance and improvement in 608 participants from 28 independent experiments. With a classification accuracy of 60% (considerably different from chance level), we identified two factors that significantly influenced neurofeedback performance: Both the inclusion of a pre-training no-feedback run before neurofeedback training and neurofeedback training of patients as compared to healthy participants were associated with better neurofeedback performance. The positive effect of pre-training no-feedback runs on neurofeedback performance might be due to the familiarization of participants with the neurofeedback setup and the mental imagery task before neurofeedback training runs. Better performance of patients as compared to healthy participants might be driven by higher motivation of patients, higher ranges for the regulation of dysfunctional brain signals, or a more extensive piloting of clinical experimental paradigms. Due to the large heterogeneity of our dataset, these findings likely generalize across neurofeedback studies, thus providing guidance for designing more efficient neurofeedback studies specifically for improving clinical neurofeedback-based interventions. To facilitate the development of data-driven recommendations for specific design details and subpopulations the field would benefit from stronger engagement in open science research practices and data sharing.

Can we predict real-time fMRI neurofeedback learning success from pretraining brain activity?

Neurofeedback training has been shown to influence behavior in healthy participants as well as to alleviate clinical symptoms in neurological, psychosomatic, and psychiatric patient populations. However, many real-time fMRI neurofeedback studies report large inter-individual differences in learning success. The factors that cause this vast variability between participants remain unknown and their identification could enhance treatment success. Thus, here we employed a meta-analytic approach including data from 24 different neurofeedback studies with a total of 401 participants, including 140 patients, to determine whether levels of activity in target brain regions during pretraining functional localizer or no-feedback runs (i.e., self-regulation in the absence of neurofeedback) could predict neurofeedback learning success. We observed a slightly positive correlation between pretraining activity levels during a functional localizer run and neurofeedback learning success, but we were not able to identify common brain-based success predictors across our diverse cohort of studies. Therefore, advances need to be made in finding robust models and measures of general neurofeedback learning, and in increasing the current study database to allow for investigating further factors that might influence neurofeedback learning.

Single session real-time fMRI neurofeedback has a lasting impact on cognitive behavioral therapy strategies

To benefit from cognitive behavioral therapy (CBT), individuals must not only learn new skills but also strategically implement them outside of session. Here, we tested a novel technique for personalizing CBT skills and facilitating their generalization to daily life. We hypothesized that showing participants the impact of specific CBT strategies on their own brain function using real-time functional magnetic imaging (rt-fMRI) neurofeedback would increase their metacognitive awareness, help them identify effective strategies, and motivate real-world use. In a within-subjects design, participants who had completed a clinical trial of a standardized course of CBT created a personal repertoire of negative autobiographical stimuli and mood regulation strategies. From each participant's repertoire, a set of experimental and control strategies were identified; only experimental strategies were practiced in the scanner. During the rt-fMRI neurofeedback session, participants used negative stimuli and strategies from their repertoire to manipulate activation in the anterior cingulate cortex, a region implicated in emotional distress. The primary outcome measures were changes in participant ratings of strategy difficulty, efficacy, and frequency of use. As predicted, ratings for unscanned control strategies were stable across observations, whereas ratings for experimental strategies changed after neurofeedback. At follow-up one month after the session, efficacy and frequency ratings for scanned strategies were predicted by neurofeedback during the rt-fMRI session. These results suggest that rt-fMRI neurofeedback created a salient and durable learning experience for patients, extending beyond the scan session to guide and motivate CBT skill use weeks later. This metacognitive approach to neurofeedback offers a promising model for increasing clinical benefits from cognitive behavioral therapy by personalizing skills and facilitating generalization.

•

Participants watched their therapy skills change brain activation in real time.

•

Rt-fMRI neurofeedback selectively predicted individual strategy efficacy ratings.

•

Ratings for unscanned control strategies did not change pre/post neurofeedback.

•

A single scan session changed metacognitions about therapy skills 4 weeks later.

Searching for IOR: Review and Results

Inhibition of return (IOR) has been proposed as an attentional mechanism which facilitates visual search by inhibiting reorienting to previously attended spatial locations. IOR is typically measured following the removal of attention from a spatial location. Early facilitation of responses to this location at early stimulus-onset asynchrony (SOA of ∼100 – 300ms) is replaced with a later and long-lasting inhibition (SOA of ∼300 – 3000ms). This inhibition has been proposed to be used by the oculomotor system to tag previously fixated locations in visual search to favour new locations over old (yet still salient) locations. Indeed, slower responses to probes presented in recently fixated locations has since been demonstrated in a variety of visual search tasks, and this form of IOR has been related to the reduced likelihood of refixating the previous or penultimate search location during natural search (MacInnes and Klein, 2003). However, recent research has challenged this interpretation by suggesting that saccadic momentum facilitates forward saccades as opposed to IOR suppressing return saccades. For instance, Smith and Henderson (2010) replicated Klein and MacInnes (1999) by finding IOR in a Where's Waldo © search task, but they reanalyzed the probability distribution of saccades to provide evidence for a saccadic momentum account. This talk will provide a review of recent research outlining the evidence for IOR and saccadic momentum in natural search patterns and present new data on the distribution of saccades in complex search tasks. [Supported by BBSRC]

Visual classification: expert knowledge guides machine learning

Humans use intuition and experience to classify everything they perceive, but only if the distinguishing patterns are visible. Machine-learning algorithms can learn class information from data sets, but the created classes' meaning isn't always clear. A proposed mixed-initiative approach combines intuitive visualizations with machine learning to tap into the strengths of human and machine classification. The use of visualizations in an expert-guided clustering technique allows the display of complex data sets in a way that allows human input into machine clustering. Test participants successfully employed this technique to classify analytic activities using behavioral observations of a creative-analysis task. The results demonstrate how visualization of the machine-learned classification can help users create more robust and intuitive categories.

Take control of quality!

Transport media were developed to maintain the viability and to prevent the overgrowth of organisms in a specimen for bacterial culture. Is what you expect what you get when you purchase transport media from a manufacturer? This article explores some of the reasons why the buyer should beware.

Management of asplenic patients in South Buckinghamshire: an audit of local practice

People without spleens have an increased risk of pneumococcal and other infections. Immunisation is advised for this group of patients, but the role of prophylactic antibiotics remains unresolved. Since 1992, general practitioners in South Buckinghamshire have been encouraged to immunise all asplenic patients against infections with Streptococcus pneumoniae, Neisseria meningitidis, and Haemophilus influenzae type b (Hib). In addition, an 'alert' card, similar in principle to a medical warning bracelet, has been produced for general practitioners to issue to asplenic patients. General practitioners' clinical records of 293 asplenic patients were subsequently examined to evaluate this programme and assess the use of prophylactic antibiotics. Uptakes of 91%, 80%, and 79% were achieved for vaccines against pneumococcal, meningococcal, and Hib infections, respectively. Twenty-three per cent of patients had been advised immediately after splenectomy to take prophylactic antibiotics. Prophylaxis was advised for different periods of time, particularly in children. Thirty-four different antibiotic regimens had been recommended for adults. Clinical records suggested that 9% of patients were taking antibiotic prophylaxis at the time of the analysis. 'Alert' cards had been distributed to 88% of patients who were eligible. It is likely that most districts within the United Kingdom could set up similar immunisation and 'alert' card programmes. The wide variation in recommendations for antibiotic prophylaxis highlights the need for further research and the development of national guidelines.

Replication of sialodacryoadenitis virus in mouse L-2 cells

Sialodacryoadenitis (SDA) is a naturally-occurring infection of the laboratory rat raused by the coronavirus, sialodacryoadenitis virus (SDAV). The study of SDAV has been limited because there is no widely available continuous cell line for the propagation of high titers of the virus. The purpose of this study, therefore, was to compare the ability of SDAV to replicate in the permanent cell lines, LBC, of rat origin, and the mouse cell lines. L-929 and L-2. Following 2 to 6 repeated passages of SDAV in LBC cells, the virus could be readily propagated in LBC and L-2 cells, but not in L-929 cells. Similarly, SDAV adapted to replicate directly in L-2 cells could be readily propagated in LBC, but not L-929 cells. In LBC and L-2 cells, cytopathic effect (CPE), viral antigen, viral particles, and virus infectivity could be demonstrated. Titers of up to 10(8.0) infectious viral particles/0.25 ml of culture fluid were obtained at 48 hours in L-2 cells. Titers in LBC cells were one to two logs lower. When susceptible rats were inoculated with eighth passage L-2 cell-adapted virus, they developed typical lesions of SDA. Virus could be recovered from infected tissues and propagated in L-2 cells on first passage. The ability to propagate SDAV to high titers in the widely available L-2 cell line should promote the study of this virus and facilitate its comparison with other murine coronaviruses.

Probing structural homologies in cell-specific glycoproteins in the leech CNS

Three monoclonal antibodies (mAbs) raised against the leech CNS recognize surface antigens on small sets and subsets of neurons or on glial cells. On immunoblots, they all recognize proteins of 130 kDa molecular weight. In addition, they each bind up to several different lower molecular weight forms. The 130 kDa polypeptides recognized by these mAbs are not major proteins on Coomassie blue-stained gels. They behave as glycoproteins on lentil lectin columns but are not major Concanavalin A-binding molecules. These molecules therefore represent a group of lower abundance, cell-type-specific antigens. Structural relationships between these antigens were explored using immunoprecipitation. The glial cell antigen was immunopurified, however, a fraction of the neuronal antigens co-precipitate. The co-precipitation of neuronal antigens raises the possibility that different neuronal antigenic determinants are carried on the same protein molecule. Such a protein may be modified to carry either one or both neuronal determinants, and could serve as a tag to physiologically delineate subsets of neurons nested within larger neuronal sets.