What if you could get rid of your fear of heights, acquire a preference for exercise, and brighten your outlook on life?

Using a combination of artificial intelligence and brain scanning technology, researchers have discovered ways of removing specific fears from the brain, increase self-confidence, and even alter personal preferences.

This technique called Decoded Neurofeedback (DecNef) can lead to the development of new interventions and treatments for patients with post-traumatic stress disorder, phobias, anxiety disorders, and other mental health conditions.

DecNef is not effective in all individuals. Understanding how the brain can regulate its own activity patterns will explain this difference in outcome and establish the technique for clinical use. To this end, DecNef researchers have released a unique dataset of five different studies in a bid to accelerate its translation from basic science to application.

The researchers, an international collaborative team led by scientists at the Computational Neuroscience Labs of the ATR Institute International in Kyoto, Japan, have released a large, open-access neuroimaging database of more than 60 individuals who underwent DecNef training. This database consists of structural and functional images of the brain, machine learning decoders, and additional processed data.

In an article titled, “The DecNef collection, fMRI data from closed-loop decoded neurofeedback experiments,” published in the journal Scientific Data, a Nature publication, the authors described the protocol they adopted in compiling the database, including common and differing scanning parameters in the source data, meta-data, structure, anonymization, cleanup, alignment, and analysis.

Decoded Neurofeedback is based on a method to read and identify specific information in the brain—for example, a fear memory.

“In Decoded Neurofeedback experiments, brain scanning is used to monitor activity in the brain, and identify complex patterns of activity that resemble a specific memory or mental state. When the pattern is detected, we give our experimental participants a small reward. The simple action of repeatedly providing a reward every time the pattern is detected modifies the original memory or mental state. Importantly, participants do not need to be aware of the patterns’ content for this to work,” said Mitsuo Kawato, PhD, director of the Computational Neuroscience Laboratories at the ATR Institute International in Japan, and senior author on the paper and who pioneered the technique a decade ago.

To date there are close to 2,000 research articles on neurofeedback, employing both functional magnetic resonance imaging (fMRI) and non-fMRI techniques, that explore the potential application of the technique in a range of conditions from autism to pain management.

“The Decoded Neurofeedback approach could have major benefits for clinical populations over traditional treatments. Patients could avoid the stress associated with exposure therapies, or side-effects resulting from established drugs. As such, it is crucial we accelerate the development of the Decoded Neurofeedback technique—and this will only be possible if more scientists are able to work on the actual data,” said Aurelio Cortese, PhD, senior researcher at ATR Institute International and lead author of the paper, explaining the vision behind the release of the data.

DecNef is a form of closed-loop fMRI neurofeedback combined with machine learning approaches. It aims to alter brain dynamics. DecNef leverages multivoxel pattern analysis (MVPA) which is why it has high target specificity.

In contrast to approaches where one measures the overall activity level within a region-of-interest, by treating each voxel (a 3D version of a 2D pixel) in isolation, MVPA is based on algorithms that learn to decode information distributed in patterns of activity.

To avoid confounding neural activity due to cognitive processing, participants in a neurofeedback experiment are not informed about the content and purpose of the experiment, or the parameters being changed.

Using a method called hyperalignment, researchers can infer the target neural representation indirectly from surrogate participants. Hyperalignment constructs a common, high-dimensional space through a set of linear transformations of patterns of neural activity across participants.

These aspects make DecNef an attractive tool to develop novel clinical applications, particularly for neuropsychiatric disorders. DecNef is also useful in studying the basic functions of the brain.

To date, DecNef has been used to investigate vision and perceptual learning in early visual cortex, subjective preference in cingulate cortex, as well as perceptual confidence in the frontoparietal network.

The DecNef database aims to foster global research in neurofeedback through meta-analyses, computational models, and neural network simulations to reveal underlying neural mechanisms. The authors believe the field will considerably benefit from having access to the dataset, which comprises five independent studies, 60 individuals, and more than 200 hours of fMRI scanning time under DecNef training.

Anyone who wants to use the dataset must apply through the ATR institutional repository or Synapse, an online repository of neuroscientific data. The details on how to access the dataset are specified in the original publication as well as on the ATR and Synapse websites.

To engage the global scientific research community in an effort to further develop the database, the authors have also released the software required to perform DecNef experiments under the terms that researchers make their data available to the scientific community through the DecNef database.