Source: Laboratories of Jonas T. Kaplan and Sarah I. Gimbel—University of Southern California
Experience shapes the brain. It is well understood that our brains are different as a result of learning. While many experience-related changes manifest themselves at the microscopic level, for example by neurochemical adjustments in the behavior of individual neurons, we may also examine anatomical changes to the structure of the brain at a macroscopic level. One famous example of this kind of change comes from the case of the London taxi drivers, who along with learning the complex routes of the city show larger volume in the hippocampus, a brain structure known to play a role in navigational memory.1
Many traditional methods of examining brain anatomy require painstaking tracing of anatomical regions of interest in order to measure their size. However, using modern neuroimaging techniques, we can now compare the anatomy of the brains across groups of people using automated algorithms. While these techniques do not avail themselves of the sophisticated knowledge that human neuroanatomists may bring to the task, they are quick, and sensitive to very small differences in anatomy. In a structural magnetic resonance image of the brain, the intensity of each volumetric pixel, or voxel, relates to the density of the gray matter in that region. For example, in a T1-weighted MRI scan, very bright voxels are found in locations where there are white matter fiber bundles, while darker voxels correspond to grey matter, where the cell bodies of neurons reside. The technique of quantifying and comparing brain structure on a voxel-by-voxel basis is called voxel-based morphometry, or VBM.2 In VBM, we first register all of the brains to a common space, smoothing over any gross differences in anatomy. We then compare the intensity values of the voxels to identify localized, small scale differences in gray matter density.
In this experiment, we will demonstrate the VBM technique by comparing the brains of musicians with those of non-musicians. Musicians engage in intense motoric, visual, and acoustic training. There is evidence from multiple sources that that the brains of people who have gone through musical training are functionally and structural different from those who haven't. Here, we follow Gaser and Shlaug3 and Bermudez et al.4 in using VBM to identify these structural differences in the brains of musicians.
1. Recruit 40 musicians and 40 non-musicians.
- Musicians should have at least 10 years of formal musical training. Training with any musical instrument is acceptable. Musicians should also be actively practicing their instrument for at least one hr/day.
- Control subjects should have little to formal training in playing a musical instrument.
- All participants should be right-handed.
- All participants should have no history of neurological, psychiatric, or cardiac disorders.
- All participants should have no metal in their bodies that they cannot remove to ensure that they are MRI-safe.
2. Pre-scan procedures
- Fill out pre-scan paperwork.
- When participants come in for their fMRI scan, have them first fill out a metal screen form to make sure they have no counter-indications for MRI, an incidental findings form giving consent for their scan to be looked at by a radiologist, and a consent form detailing the risks and benefits of the study.
- Prepare participants to go in the scanner by removing all metal from their body, including belts, wallets, phones, hair clips, coins, and all jewelry.
3. Put the participant in the scanner.
- Give the participant ear plugs to protect their ears from the noise of the scanner and ear phones to wear so they can hear the experimenter during the scan, and have them lie down on the bed with their head in the coil.
- Give the participant the emergency squeeze ball and instruct them to squeeze it in case of emergency during the scan.
- Use foam pads to secure the participants head in the coil to avoid excess movement during the scan, and remind the participant that it is very important to stay as still as possible during the scan, as even the smallest movements blur the images.
4. Data collection
- Collect a high-resolution anatomical scan of the whole brain. This scan should be a T1-weighted sequence such as a Magnetization Prepared Rapid Gradient Echo (MP-RAGE) with isotropic 1 mm voxels.
5. Data analysis
- Remove the skull from each anatomical brain scan using automated software. Check the skull stripping for quality.
- Create a study-specific gray matter template using an iterative process of linear and non-linear registration (Figure 1).
- Use automated software to segment each subject's brain into white matter, gray matter, and CSF based on the intensity at each voxel.
- Perform a linear affine transformation with 12 degrees of freedom to register each subject's brain to a standard atlas space, such as the MNI152 atlas.
- Warp each subject's gray matter image into this space, and average them all together.
- Mirror this image left to right and then average the mirror images together to produce the gray matter template brain.
- Re-register each subject's brain to this template using a nonlinear transformation.
- Average all of the newly transformed brains together.
- Make a mirror image of this new template, and average the two mirror images together to produce a final gray matter template for this study.
Figure 1: Creation of study-specific gray matter template. Using iterative linear and nonlinear transformations, each brain is registered to a common space and averaged together to create a study-specific gray matter template brain.
- Register each subject's gray matter image to the template and preprocess.
- Use a nonlinear transformation to register each subject's brain to the study-specific template.
- To compensate for the amount each brain structure has been stretched to fit the template space, multiply by a measure of how much warping has been done. This measure is called the Jacobian of the warp field. This step is to account for the fact that structures may appear to have more gray matter simply because they have been stretched more by the nonlinear registration process.
- Smooth the data using a Gaussian kernel with a Full Width Half Maximum (FWHM) of 10 mm.
- These aligned, smoothed brains will serve as the final data for voxel-based analysis.
- Use the General Linear Model to analyze the difference between the groups at every voxel.
- Model each group of brains with a separate regressor, and compute a contrast that compares the two groups, generating statistical maps that quantify the likelihood of differences at each voxel.
- Threshold the statistical maps to identify statistically significant clusters.
- Employ a multiple comparisons correction technique such as False Discovery Rate (FDR) to control for the fact that were are doing thousands of simultaneous statistical tests. With FDR, a q value of 0.01 will estimate the rate of false positives above threshold to be 1%.
Our brains are shaped by experiences, resulting in changes in cortical volume.
For instance, certain proficiencies, like learning and mastering a second language, have been shown to increase the density of gray matter, where cell bodies reside, particularly in structures such as the frontal lobe.
Before modern advances, to measure a particular area’s size, scientists would have to painstakingly trace the region of interest—a very tedious task. Now, more sensitive neuroimaging techniques—known as voxel-based morphometry, VBM—exist to capture small volumetric differences in neuroanatomy.
Based on previous work of Gaser and Shlaug, as well as Bermudez and colleagues, this video demonstrates how to collect structural magnetic resonance images and use VBM to identify the intensity values of voxels in the brains of individuals with different experiences—expert musicians compared to those with very limited training—as well as in other cases of expertise, such as chess playing.
In this experiment, two groups of participants—formally trained musicians and controls with no such training—are asked to lie in an MRI scanner while structural images of their brains are collected.
Particular regions can then be defined using an automated approach, based on the intensity of volumetric pixels, called voxels. For instance, very bright clusters indicate the location of white-matter fiber bundles, while darker voxels correspond to areas with dense gray matter.
Following this segmentation for each brain, the images are transformed—registered to a standard atlas, which is a common space to allow for between-subject comparisons.
Often times, this registration process can stretch out an image, which makes some structures seem like they have more gray matter than they really do.
Therefore, the template must be multiplied by a measure of how much warping has been done, called a Jacobian determinant, to compensate for the repeated stretching, and then all gross differences in anatomy are smoothed out.
After the transformations are applied, the dependent variable is calculated as the differences in gray matter density between musicians’ brains compared to non-musician controls.
Due to the increased use of complex auditory processing in skillful musicians, it is expected that this group will show increased gray matter density in auditory brain regions, such as the superior temporal lobe and Heschl’s gyrus, compared to the control group.
Prior to the experiment, recruit 40 musicians who actively practice any instrument 1 hr a day and have at least 10 years of formal musical training, as well as 40 non-musician controls who have little to no proper training.
On the day of their scan, greet each participant in the laboratory and verify that they meet the safety requirements as they complete the necessary consent forms.
Please refer to another fMRI project in this collection for more details on how to prepare individuals to enter the scanning room and scanner bore.
Now, instruct the participant to lie still in the scanner, and begin scanning the whole brain by collecting a high-resolution, T1-weighted anatomical sequence such as Magnetization Prepared-Rapid Gradient Echo with 1 mm isotropic voxels.
Following the image-collection protocol, dismiss the participant and start the analysis.
To begin preprocessing, isolate the brain from the skull for each scan and check the quality of the stripping.
For this study, create a specific gray matter template by first segmenting each subject’s brain into white and gray matter and cerebral spinal fluid, CSF, based on the intensity of each voxel. Note, the software automatically distinguishes bright voxels as white matter, dark voxels as gray matter, and areas within the ventricles as CSF.
Perform a linear affine transformation with 12° of freedom, to register each subject’s brain to a standard atlas space. Warp each subject’s gray matter image into this space, and average them all together.
Next, mirror this left to right, and once again, average the images together to produce the initial gray matter template.
Then, perform a non-linear transformation to re-register each subject’s brain to the gray matter figure, and average these together. Create a mirrored copy of this new image, and once again average the two together to produce a final, study-specific, gray matter template.
Now register each subject’s brain to the last gray matter figure using a non-linear transformation, and multiply by a Jacobian measure of how much warping has been done to compensate for the amount each brain structure has been stretched to fit the template space.
Subsequently, smooth the data using a Gaussian kernel with a Full-Width Half Maximum of 10 mm to increase the overlap of similar brain voxels across all subjects.
With preprocessing completed, model each group of brains with a separate regressor. Compute a contrast that compares the two groups to generate statistical maps that quantify the likelihood of differences at each voxel.
Finally, perform a multiple comparisons correction technique, such as a False Discovery Rate with a q value of 0.01, to control for the thousands of simultaneous statistical tests performed. This value will estimate the rate of false positives above a threshold of 1%.
Here, the VBM analysis revealed significant bilateral increases in gray matter density in the superior temporal lobe of the musicians’ brains compared to the controls. The greatest difference was shown on the right side, and this included the posterior portion of Heschl’s gyrus, the location of the primary auditory cortex.
Now that you are familiar with how to use VBM to study neuroanatomy, let’s look how researchers use this technique to study structural differences in other populations.
While many tasks involving intense training and experience are associated with increases in gray matter volume, this enlargement is not always the case for all types of learned skill-sets, like in the brain of an experienced chess player.
When compared to controls, gray matter volume was reduced in the occipito-temporal junction, an area important for object recognition. Such findings result in an interesting anomaly that may help scientists further understand how cortical volume relates to performance in demanding tasks.
Individuals who are blind from birth often have smaller gray matter volume in their visual cortex compared to controls. Interestingly, through the use of VBM, researchers have discovered significant enlargement in areas of the brain not responsible for vision, such as the auditory cortex, which was twice the size found in sighted controls.
These structural differences may serve as an anatomical foundation to explain why other senses are heightened in blind individuals.
Furthermore, structural MRI and VBM analysis on medication-naïve patients with major depressive disorder also indicate differences in gray matter volumes compared to controls.
Scientists found that these patients had decreased gray matter volume in the frontal cortex and insula, which may explain why depressed patients have difficulty with cognitive control over negative feelings toward themselves and others.
You’ve just watched JoVE’s video on voxel-based morphometry. Now you should have a good understanding of how to collect anatomical images using MRI as well as how to analyze and interpret differences in gray matter intensity in regions of the auditory cortex. You should have also learned that not all areas of expertise lead to increases in cortical density.
Thanks for watching!
The VBM analysis revealed significant localized increases in gray matter density in musicians' brains compared with non-musician controls. These differences were found in the superior temporal lobes on both sides. The largest, most significant cluster was on the right side and includes the posterior portion of Heschl's gyrus (Figure 2). Heschl's gyrus is the location of the primary auditory cortex, and the surrounding cortices are involved in complex auditory processing. Thus, these results are consistent with previous findings of morphological differences between musicians and non-musicians in auditory brain regions.
Figure 2: Gray matter differences between groups. Musicians showed significantly higher gray matter density in the superior temporal lobe on both sides, with the greatest differences on the right side. This region includes part of Heschl's gyrus, the primary auditory cortex.
Applications and Summary
The VBM technique has the potential to demonstrate localized differences in gray matter between groups of people, or in association with a measurement that varies across a group of people. In addition to finding structural differences that relate to different forms of training, this technique may reveal anatomical differences that are associated with wide ranging neuropsychological conditions such as depression,5 dyslexia,6 or schizophrenia.7
It is important to note that there are multiple explanations for the existence of between-group differences in brain anatomy. For example, in the case of musicians, there could be a self-selection bias. We may find such differences if people with a certain brain anatomy are more likely to become musicians. In order to establish that structural differences between groups of people are the result of experience, the most definitive way is to employ a longitudinal study that follows people over time.
- Maguire, E.A., et al. Navigation-related structural change in the hippocampi of taxi drivers. Proc Natl Acad Sci U S A 97, 4398-4403 (2000).
- Ashburner, J. & Friston, K.J. Voxel-based morphometry--the methods. Neuroimage 11, 805-821 (2000).
- Gaser, C. & Schlaug, G. Brain structures differ between musicians and non-musicians. J Neurosci 23, 9240-9245 (2003).
- Bermudez, P., Lerch, J.P., Evans, A.C. & Zatorre, R.J. Neuroanatomical correlates of musicianship as revealed by cortical thickness and voxel-based morphometry. Cereb Cortex 19, 1583-1596 (2009).
- Bora, E., Fornito, A., Pantelis, C. & Yucel, M. Gray matter abnormalities in Major Depressive Disorder: a meta-analysis of voxel based morphometry studies. J Affect Disord 138, 9-18 (2012).
- Richlan, F., Kronbichler, M. & Wimmer, H. Structural abnormalities in the dyslexic brain: a meta-analysis of voxel-based morphometry studies. Hum Brain Mapp 34, 3055-3065 (2013).
- Zhang, T. & Davatzikos, C. Optimally-Discriminative Voxel-Based Morphometry significantly increases the ability to detect group differences in schizophrenia, mild cognitive impairment, and Alzheimer's disease. Neuroimage 79, 94-110 (2013).