Why functional MRI is the gateway to better diagnosing our patients.
Since the late 1800s, researchers specializing in neuroscience have made an effort to image the structure or function of the nervous system. Developments such as the human circulation balance invented by Angelo Mosso in the 1880s, pneumoencephalography in the early 1900s, and basic computed tomography in the late 1900s have helped scientists formulate the ultimate form of brain imaging: functional magnetic resonance imaging (fMRI).
In order to comprehend how fMRI works, it is important to understand how MRI scanners operate in general. All MRI techniques utilize magnetic fields to manipulate the hydrogen atoms within the brain. The atoms are then met with radio waves, which force the atoms to ‘spin’. This spin is recorded and the computer software develops images reflecting the information collected by the scanner, specifically the levels of tissue density. The technique of fMRI is characterized as ‘functional’ because it measures brain functionality as it performs a specific task, unlike other forms of static MRI.
How Is fMRI Different?
Functional MRI is unique, however, in that it measures something different. fMRI machines are known to engage in blood oxygen level dependent (BOLD) imaging, which, simply put, analyzes the varying levels of oxygenated blood throughout the brain. The reasoning for this is that certain tasks performed by the brain cause heightened levels of blood flow to that particular area whereas the areas of the brain that do not experience any stimulation don’t behave the same way.
So, how does this relate to the magnetic properties of MRIs? Well, oxygenated and deoxygenated blood (normally referred to as oxyhemoglobin and deoxyhemoglobin) have different concentrations of hydrogen, and thus, exhibit distinct magnetic properties. This difference is what the computational software capitalizes on; the disparity between the respective levels of oxy- and deoxyhemoglobin causes a subtraction to occur. This process is most common in activation studies, which link neural responses to experimental tasks using a statistical approach. Generally, activation cases detect differences in brain activity across several stimuli or across rest and active states during a specific experimental task. Functional imaging utilizes both activation studies, which capture localized areas of activation, and functional connectivity studies, which capture activation across regions or lobes of the brain that respond to the same experimental task or stimuli.
Analyzing My Own Scans
This lab took things into a computational perspective; their goal was not so much to study patient brains, but primarily to use patient scans to develop more effective scanning techniques and processing methods. As such, the researchers there were experts in experimenting with fMRI, so it was a perfect fit for me. I figured that the best way to grasp the concept of functional imaging was to be a part of the process from start to end. So naturally, I volunteered to get scanned by the MRI. For details on how the scan went, what I had to do inside, and what to expect, read my other post specifically about my MRI experience here.
During my scan, I performed four separate tasks. The first task was a hand motor task in which I had to perform an action with my hand when prompted. The later three tasks were language tasks, one sentence completion task, in which I filled in a blank to complete a sentence, on verb generation task, in which I thought of a verb that best matched a random noun, and lastly, one word generation task, during which I imagined a word that started with the letter shown to me. Again, I go into more detail in my other post dedicated to my MRI experience and what others should expect.
Before analyzing my scans, I made several predictions. I hypothesized that my hand motor task would activate my primary motor cortex along the motor strip at the top of my brain. I also predicted that the activation would show up bilaterally, since using both my right and left hands triggers both hemispheres of my brain. I also considered the possibility that my occipital lobe would light up as well, since the task requires the patient to see the screen. However, at the time of my scan, I was not wearing my glasses, so I changed my mind and predicted that my occipital lobe would show minimal activation due to less visual concentration due to slight squinting. With regards to the other three language tasks, I was hypothesizing activation in the Broca and Wernicke’s regions of the brain located medially around the temporal lobe. The Broca and Wernicke’s areas are unique to language comprehension. The word and verb generation tasks have different onsets and are a different length in terms of the number of rest and active phases, but this would only affect the analysis process and not particularly where the brain would be activated. Finally, I acknowledged the respective Brodmann areas that should be ideally activated. Brodmann areas are regions of the cerebral cortex that are defined by their cellular makeup and neural specialization. I predicted that Area 4 (primary motor cortex), Area 6 (coordination), Area 7 (visuomotor coordination), Area 18 (interpretation of images) and Area 17 (primary visual cortex) would be activated for the hand motor task. For the language tasks, Brodmann Areas 22, 39, 40 (Wernicke’s area) and Area 44 (Broca’s area) should be the main areas of activation. Figure 1 and 2 depict the Brodmann areas that were mentioned within circles.
A common misconception is that MRI machines can directly illustrate brain activity. This is not the case, as the MRI output is simply raw data that can only become meaningful once it is processed through a series of steps. Certain statistical techniques are required in order to derive the correlation between stimuli and task and to properly use neuroimaging data for diagnostics. This is where the computational aspect of computational neuroscience comes into play. In order to analyze the data set, I used the statistical parametric mapping (SPM) method, which refers to the spatially extended statistical process used to read functional imaging data.
Below, I outlined the general steps that I took in order to process my brain scans. These steps are more complex then they come across, but they briefly summarize the purpose(s) of each.
The preprocessing steps are simply a form of preparation. Specifically, preprocessing helps to eliminate excess noise, distortions, or disorientation. The term ‘noise’ refers to inaccurate data that is created along with the data that is desired and can come from various movements by the patient inside the scanner. If these steps were to be skipped, the final activation maps could become inaccurate. For example, there could be activation shown in the skull area where brain activity is obviously impossible.
- Reorientation & Realignment: To avoid the misalignment of brain scans across multiple experimental tasks, all converted functional volumes were realigned with correspondence to one image (usually the first displayed functional volume). They are also estimated and resliced so that all slices of the image correspond to the same point.
- Coregistration: Once the data is preprocessed, it can be overlaid on the anatomical brain images of the patient.
- Segmentation: Segmentation is not as widely popular a step as the others, but Jefferson utilizes it in the data processing method to combine spatial normalization and save bias corrected.
- Normalization: Spatial normalization helps bind scans to a standard template, which is usually from the Montreal Neurological Institute. Normalization makes it easier to compare activity.
- Smoothing: Spatial smoothing is a process that entwines the functional volumes and increases the signal-to-noise ratio. The idea is that smoothing can subdue neuroanatomic variation and allows us to threshold easier.
- Display & Statistical Analysis: The final step includes the statistical analysis using voxel-based analysis and display of the brain using overlays or the maps in the axial or sagittal view.
In these steps, I did not go into detail on voxel (3D pixel) sizes, onsets, durations, thresholds, or mini steps along the way. Some of these vary from test to test.
What Do the Results Say?
For the hand motor task, my prediction was correct in that the primary motor cortex did light up bilaterally. I was also correct about my vision playing a role in the lack of activation in the occipital lobe. Figure 3 depicts activation through the axial, coronal, and sagittal views with a scale provided.
The sentence completion activation maps contained a lot of false positives. I concluded that there was an error in the postprocessing that caused the maps to turn out inaccurately. There is no clear activation in either of the language areas, and all activation appears to look uniform. This is not what a ‘normal’ activation map should look like, but it is provided for the sake of transparency and comparison in Figure 4. These are the full activation maps in both the sagittal and axial views.
My predictions for the verb generation map were similar to the outcome. In the verb generation activation map, there is evident activation in the Broca area, and although there is minimal Wernicke activation, there is still color in the area. The activation map in sagittal view has been provided in Figure 5. Figure 5.1 depicts the activation on a coordinate plane, with the coordinates listed from highest activation to lowest below and the rest/active duration cycle to the right. This image is from right before the statistical analysis was complete and the activation map was generated.
The results for the word generation maps were similar to those of the verb generation. After all, they are nearly identical tests that measure nearly the same brain activity. The maps show strong activation in the Broca region and some in the Wernicke’s region in both the sagittal and axial views. The voxel size for all of these scans was 3 mm, and the extent threshold was at 0.05, with the thickness altered to 5 instead of 1. Figures 6 and 6.1 are shown below.
In short, the activation maps for my four separate tasks were close to the ideal levels of activation with the exception of the sentence completion map. The maps depicted activation in the previously mentioned Brodmann areas and accurately reflected the task that was performed. Since these tests were simple in nature and were not specialized for a test for a condition or disease, it is not possible to diagnose any illnesses from these maps alone. What these maps do showcase, however, is that my brain is normal in behavior with regard to the specific function the tasks test for. In other words, my bodily motor functions and language comprehension are healthy. It is important to remember that not all activation maps will look alike amongst patients. One patient may display strong activation in an area where another may show weak activation for the same task. Some patients even display different areas of activation and can still be considered healthy.
Applications & Outlook
The ultimate goal of fMRI imaging is to generate accurate brain maps that depict the activity in the brain in a non-invasive manner. These activations can be manipulated to provide more telling scans and more useful maps. It is widely understood that this type of imaging can not only shed light on how our cognition and emotions affect our behaviors, but also that it can uncover the basis of neurological disorders. Imaging techniques are quickly developing and providing quicker, easier, and more informative views of the human brain. Thus far, MRIs have had a great impact on medical practices, and scientists believe that adding a functional aspect changes everything.
Functional MRI is still in a learning stage of clinical use, and it has a long way to go. Due to its young age, scientists are still setting standards. Until then, fMRI scanners will be experimented with like toys in labs around the world, with researches hoping that they will be the mark of something extraordinary in the discipline of neuroscience.
These sorts of developments in neuroimaging may not eliminate all of our neural inquiries, but they do bring us one step closer to a healthier world.
Published 1:55 AM EST