Reading and controlling human brain activation using real-time functional magnetic resonance imaging (rt-fMRI)

Reading and controlling human brain activation using real-time functional magnetic resonance imaging
R. Christopher deCharms
Omneuron 3T MRI Research Center, Menlo Park, CA 94025, USA
TRENDS in Cognitive Sciences Vol.11 No.11

Understanding how to control how the brain’s functioning mediates mental experience and the brain’s processing to alter cognition or disease are central projects of cognitive and neural science. The advent of realtime functional magnetic resonance imaging (rtfMRI) now makes it possible to observe the biology of one’s own brain while thinking, feeling and acting. Recent evidence suggests that people can learn to control brain
activation in localized regions, with corresponding changes in their mental operations, by observing information
from their brain while inside an MRI scanner. For example, subjects can learn to deliberately control activation in brain regions involved in pain processing with corresponding changes in experienced pain. This may provide a novel, non-invasive means of observing and controlling brain function, potentially altering cognitive processes or disease.

Personal notes:

How does the brain mediate mental experience?
How does the brains processing alter cognition or disease?

Both are projects central to cognitive and neural sciences.

rtFMRI: allows you to observe the biology of your brain while thinking, feeling and acting. You can learn how to control brain activation in localized regions in the brain with corresponding changes in mental operations.
example: people can learn how to control activation in brain regions involved in pain processing with a corresponding change in the perception of pain.

People control brain activation all the time: every voluntary action engages a specific brain mechanism. rtfMRI has the ability to allow people to see the brain processes that underlie their current thoughts and feelings and to change them.

So, brining brain activation into awareness may allow us to learn greater explicit control over our own cognitive and neural states or to mimic the states of others.

What are the limits of learned control?
What brain areas can be brought under conscious control?
What cognitive or neural states can be deliberately learned or modified?
Is it possible to target neural plasticity to enhance function in particular brain systems through training?
Other important application of rtFMRI:
monitoring experiments, improving quality assurance, data acquisition, rapid mapping for surgical planning purposes.
Learning control over brain activation, 2 stages:
1. Red out step: method for reading out brain processes. Persons current brain activation (single area, multiple area or spatial pattern) can be compared with another person, a large group of people or with a hypothesized activation pattern.
2. Training step: method for training a person to use the readout information to control and manipulate those regions. learn to select, refine or create new cognitive strategies or neural processes to optimize activation in the region of interest (ROI).

Step 1: Reading out brain states in rtfMRI

A. rtfMRI methodology and development

Massive parallel measurement of brain physiology; 2^16 spatial locations as 3D stacks of images every 1-2 seconds.
Developed in 1995 by cox et al. but application was not readily recognized.
Limitations of the rtFMRI signal.
Limitation in spatial and temporal resolution and noise.
Temporal limitation: Neurons fire at the scale of milliseconds and precise timing is known to be an important in neuronal coding. fMRI signal is derived from blood and oxygenation levels 2-6 seconds after neural activation and rtFMRI adds another 1-4 seconds of delay.
Temporal limitation: fMRI is measured from voxels (3Dpixel) on the order of 3x3x3mm, whereas neurons have micron spatial scale features.
Noise: MRI signal is inherently noisy. Raw fMRI images have a low signal to noise ratio, blurring, spatial distortion, signal loss and drifts in signal intensity over time.
Physiological noise: physiology adds other artifacts as a result of fluctuations in blood flow, respiration, heart rate, neuronal signals, motion.
This made it unclear whether the signal can be used to interpret continuously evolving brain functions.
B. rtFMRI pattern classification for reading brain activation.

Pattern classification constitutes a set of mathematical methods that compare two or more spatial or temporal patterns of data for the purpose of assessing similarity. Pattern classification algorithms represent an important new method for reading brain states in real time.
Reference patterns are usually created for a specific brain pattern. How?
The spatial pattern of fMRI activation for voxels from selected brain regions during a particular task are compared with voxels from a rest period when the subject performs an overt or unrelated task. A value of activation (decrease or increase) is recorded and compared quantitative with a specific reference pattern or multiple patterns utilizing a pattern classification algorithm.
Given that rtfMRI has a large number of parallel spatial measurements, pattern recognition algorithms can be used to extract greater information from individual ROIs. Pattern classifiers may be built automatically by providing many samples of actual data from a specific brain state; the pattern classifier will seek to find statistical similarities from data within each state.

Pattern classifier algorithms have been used to discern whether a subject is thinking happy or sad thoughts or thinking happy or sad thoughts. however, more research is needed to tell what information real time classifiers can be extracted. This is important since we can train subjects to produce or control arbitrary complex patterns of brain activation learning to produce desirable brain states through mimicking others.

Step 2: Learned control over brain activation

A. Prior methods for physiological self-regulation

It has been previously shown that we can learn volitional control over autonomic measures (heart rate, skin conductance, muscle tone) and measures of brain activation including EEG rhythms, slow cortical potentials, and single motor neurons firing in prosthetic control.
If subjects can learn control over autonomic measures and EEG rhythms, can they learn to volitional control over highly specialized neurophysiological functions mediated by hundreds of brain areas?
rtfMRI: Precise localization is critical for learning control over a brain region and its highly specific functions.
EEG and MEG have a particular difficulty localizing activation sources in the brain from continuously acquired data.
Advantage of EEG: portable, inexpensive, provides precise temporal resolution.
Principles of learning precise control over the motor cortex using fMRI may be more analogous to principles of learning to play the piano than to learning relaxation using EEG feedback or heart rate conditioning.

B. rtFMRI-based training methods.

Basically you are performing a brain imaging experiment on yourself trying to optimize a desired activation pattern by selecting cognitive and neural processes using feedback. The goal is to enhance control over brain activation corresponding with an enhancement in control over a related cognitive process.
However, do changes in brain activation arise as a result of changes in cognitive processes; or do changes in brain activation cause changes cognitive processes.
Typical Protocol: subjects are instructed to learn in increase or decrease activation in an ROI during alternating time period. success is presented as brain activation data in real time. Movement, respiratory changes, arousal present artifacts that must be avoided during the experiment.

C. Training control over brain activation using real time or near real time fMRI feedback.

Several studies have demonstrated that the fMRI signal is sufficiently statistically robust to serve as a meaningful basis for training.
Advantage: signal may be averaged to provide more statistically robust information. Disadvantage: a significant delay is introduced before the subject receives the information and removes temporal information contained in ongoing signal changes.
Anterior cingalate ucoretex (ACC): first region where a subject learned control over activation using rtfMRI feedback.
deCharms et al.(2004 study) showed that repeated training can teach subjects to achive greater control over areas in the somatosensory cortex (tactile, movement, motor mental imagery).
control group was shown sham fMRI feedback. this group did not show improved control over brain activation, demonstrating that learned control was due to training using rtfMRI. When rtfMRI information was withdrawn from subjects who learned control, they could still exercise the same control.
Wiki_ Motor mental imagery: a visio-motor mental process by which an individual rehearses or stimulates a given action.
Limitation of rtFMRI: small signal changes may never be detected by a subject (example: subtle change in BOLD), whereas a large signal change is almost immediately apparent (dramatic response due to finer tapping). Data is temporally filtered to remove noise and slow drifts, leading to an implied temporal resolution on the order of 1 to 5 seconds.

D. Brain aras subject to volitional control

Brain Cartography
What other brain areas can we learn to control? What type of tasks of stimuli are required to teach control and what behavioral consequences result from this control?
This is a rich area for further investigation.
Initially, brain areas that were extensively studies, easy to image, and presumed readily controllable based upon function were targeted for control over activation.
These readily controllable areas include the somatosensory cortex, the parahippocampal place area (PPA), the amygdala, the auditory cortex, the insular cortex, the Anterior cigualte corxtex.

Future Directions

A. rtfMRI training of multiple brain regions as a brain computer interface (BCI)
Can we control and train more that one brain regions simultaneously?
Studies have shown that we can differentially control two brain areas.
subjects were provided rtfMRI feedback on the difference between the supplementary motor area (SMA) and the PPA, and were asked to use visual and motor imagery to activate these two structures. Results showed an increase in the differential signal between SMA and PPA.
example 2:
subjects were able to navigate through a 2d maze by using mental calculations to activate the medical superior frontal and anterior cingulate gyri, left and right mental imagery to activate somatosensory areas, and mental speech generation to activate the posterior superior temporal gyrus.

B. rtfMRI training leading to changes in cognition or behavior
Learned control over brain activation may lead to changes in cognition, behavior or disease processes. Example: Pain perception change rACC study by deCharmes.

C. The potential for clinical Applications

rtFMRI has the potential to be developed into a clinically important tool that is non-invasive, non-pharmacologic, and potentially safe and reversible.
Exercise of brain systems through rtfMRI training could lead to long-term up regulation in brain functions based upon use-dependent plasticity.
their is potential for treatment through localized control through brain activation in chronic pain, control over addiction and craving, depression, recovery from stroke, and as an adjunct to psychotherapy.
Additionally, rtFMRI may provide new ways for diagnosis, guiding therapeutic interventions, and to assess the mechanisms of efficacy of treatments.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

  • Top Rated

  • العقل السليم في الجسم السليم

  • Enter your email address to subscribe to this blog and receive notifications of new posts by email.

    Join 31 other followers

  • Meta

  • Blog Stats

    • 131,828 hits
%d bloggers like this: