104
Y. Zhang
imaging, it will first be necessary to understand the nature of 4D magnetic resonance
imaging; fMRI is a complex imaging modality that requires some investigation to
appreciate, especially when contrasted with the more direct and simple EEG detection. For this, we will not be focused on the actual electrical activity of the brain, but
will instead address the blood circulating within the cortical tissues.
5.3.2.1 The Basics of FMRI Acquisition and Analysis
When neurons fire within the brain, large amounts of energy are necessarily consumed
by cells in the process of regulating and restoring their ionic gradients. The energy for
this is provided in the form of blood borne glucose and oxygen, whose consumption
creates a local decrease in oxy-hemoglobin. This decrease, paired with the local
release of neurotransmitters, is perceived by astrocytes in the area of activity. The
astrocytes then alter the blood flow of the brain, creating a sharp influx of oxygenated
to compensate for the deficit caused by functional activity. This association between
neural activity and cerebral blood flow is known as neurovascular coupling (NVC).
The compensatory increase in local oxygen then serves as the basis for fMRI imaging.
fMRI takes advantage of this coupling, along with the fact that oxygenated and
deoxygenated blood present with very different magnetic properties, to provide an
indirect depiction of cortical activity. A series of magnetic fields are applied and
pulsed across the sample or subject to rotate and displace the dipole moments of
blood-borne hemoglobin. As these displaced dipoles return to their original states,
the extra energy imparted by the magnetic field is released as a radiofrequency signal
that travels unhindered through the skull, scalp, and open air. The detection of this
signal is interpreted to produce greyscale images of the whole brain, one slice at a
time. Statistical analyses are typically applied to identify which specific regions (in
the form of voxels) show statistically significant differences between conditions. The
result of this process is an image that is slow and indirect, but very spatially accurate.
This forms a complimentary imaging modality to EEG, which is both fast and direct
in its depiction of cortical activity but struggles with spatial blurring and inaccuracy.
While it is not considered a primary focus of interest, it will be worthwhile to
spend some extra time discussing the primary statistical methods of fMRI. When
MRI data is acquired, it comes in the form of greyscale 3D volumes. Anatomical
scans will consist of a single high resolution volume while functional scans will
consist of multiple lower resolution images taken at each timepoint during the scan.
Each image will be made up of a series of 3D voxels, which are analogous to the
pixels measured on 2D displays. Over the course of the scan, the intensity of each
voxel will fluctuate both due to noise and actual shifts in the magnetic properties
of the underlying tissues. A general linear model (GLM) is then constructed to
determine the statistically significant changes in greyscale value that can be attributed
to changes in the experimental condition. Following the simple, ideal form of the
GLM, the intensity of a voxel (Y) should be linearly related to the condition (X)
following the simple formula:
Y. Zhang
imaging, it will first be necessary to understand the nature of 4D magnetic resonance
imaging; fMRI is a complex imaging modality that requires some investigation to
appreciate, especially when contrasted with the more direct and simple EEG detection. For this, we will not be focused on the actual electrical activity of the brain, but
will instead address the blood circulating within the cortical tissues.
5.3.2.1 The Basics of FMRI Acquisition and Analysis
When neurons fire within the brain, large amounts of energy are necessarily consumed
by cells in the process of regulating and restoring their ionic gradients. The energy for
this is provided in the form of blood borne glucose and oxygen, whose consumption
creates a local decrease in oxy-hemoglobin. This decrease, paired with the local
release of neurotransmitters, is perceived by astrocytes in the area of activity. The
astrocytes then alter the blood flow of the brain, creating a sharp influx of oxygenated
to compensate for the deficit caused by functional activity. This association between
neural activity and cerebral blood flow is known as neurovascular coupling (NVC).
The compensatory increase in local oxygen then serves as the basis for fMRI imaging.
fMRI takes advantage of this coupling, along with the fact that oxygenated and
deoxygenated blood present with very different magnetic properties, to provide an
indirect depiction of cortical activity. A series of magnetic fields are applied and
pulsed across the sample or subject to rotate and displace the dipole moments of
blood-borne hemoglobin. As these displaced dipoles return to their original states,
the extra energy imparted by the magnetic field is released as a radiofrequency signal
that travels unhindered through the skull, scalp, and open air. The detection of this
signal is interpreted to produce greyscale images of the whole brain, one slice at a
time. Statistical analyses are typically applied to identify which specific regions (in
the form of voxels) show statistically significant differences between conditions. The
result of this process is an image that is slow and indirect, but very spatially accurate.
This forms a complimentary imaging modality to EEG, which is both fast and direct
in its depiction of cortical activity but struggles with spatial blurring and inaccuracy.
While it is not considered a primary focus of interest, it will be worthwhile to
spend some extra time discussing the primary statistical methods of fMRI. When
MRI data is acquired, it comes in the form of greyscale 3D volumes. Anatomical
scans will consist of a single high resolution volume while functional scans will
consist of multiple lower resolution images taken at each timepoint during the scan.
Each image will be made up of a series of 3D voxels, which are analogous to the
pixels measured on 2D displays. Over the course of the scan, the intensity of each
voxel will fluctuate both due to noise and actual shifts in the magnetic properties
of the underlying tissues. A general linear model (GLM) is then constructed to
determine the statistically significant changes in greyscale value that can be attributed
to changes in the experimental condition. Following the simple, ideal form of the
GLM, the intensity of a voxel (Y) should be linearly related to the condition (X)
following the simple formula: