fMRI Explained (Functional Magnetic Resonance Imaging) transcript

These types of blobs and these and these and these are probably distantly familiar to you as something to do with brain activity. But w ab what are the blobs? I'm going to explain exactly what these red blobs are and what they represent hopefully so you have an intuitive understanding of how functional MRI fris and what it's measuring. I've now completed my PhD in functional MRI and this is the video I wish I had been able to watch on my very first day.

This is an MRI machine, magnetic resonance imaging. The physics of MRI is insanely complicated and I am no physicist. But the physics is not actually necessary to understand what these blobs are. To give you the key idea though, when you enter an MRI machine, a significant number of the hydrogens in the water molecules in your body all align in the same direction. Radio frequency pulses are then sent into your body which causes a radio signal to be released from the hydrogen which the MRI machine detects.

This is similar to clapping your hand above guitar strings, then actually faintly hearing those guitar strings. You will hear their resonance. I don't know if this is going to work. [applause] An MRI scan is doing a similar thing to your brain. The radio signals or resonance that is detected are computationally converted into an image. Beautiful images like this, this and this. There are two key types of MRI scan, both from exactly the same MRI machine, just different software running the scans.

One is called structural MRI. This is the camera mode and takes one detailed picture of the entire brain, which usually takes about 5 to 10 minutes just for one entire brain volume. This is your breadandut diagnostic test to see if there's any abnormalities like a tumor or a bleed or as all participants joke, absolutely nothing at all in their brain. Let's put this nice highquality image of a structural MRI scan here which took just under 10 minutes to acquire.

Another type of scan is our functional MRI scan. This is the video camera mode and takes multiple images of the brain very quickly, each about 2.4 seconds. This is the repetition time or TR for the extra nerdy among you. Let's put a functional scan of the same brain that took 2.4 seconds to acquire here. Notice how the functional scan is a bit more blurry and has worse resolution than the structural scan because it was acquired 250 times more quickly than the structural scan.

Anyways, this is our first fMRI brain volume. Then 2.4 seconds later, another. Then 2.4 seconds later, another. Then 2.4 seconds later, another. Then 2.4 seconds later, another. Then 2.4 seconds later, another Then 2.4 seconds later, another. Then 2.4 24 seconds later, another. So, as you can see, we start to get information about what's going on inside the brain every couple of seconds across a period of time. It's not actually brain activity itself we measure, as you're about to see.

But this is the time [snorts] dimension of fMRI that makes it such a powerful tool for measuring brain activity. This is also a great time for me to introduce or remind you of this word, a voxil. This is a voxil, but you can't really see it here. Let's zoom in. H, still not really. Let's zoom in again. Okay, cool. Now we can see the voxil. A voxil is the smallest 3D volume represented in an MRI image which contains a single measured signal value.

So looking at our image here, this is a voxil and this is a voxil and this is a voxil and this is a voxil etc. A single voxil is just like a 3D pixel. In the case of this image, each pixel or voxil is 3 mm by 3 mm by 3 mm. And remember, inside the 3 mm cube placed anywhere in the brain, you'll have millions of neurons. So, we aren't looking at anywhere near individual neuronal activity, only clusters of millions of neurons.

So, we have our fMRI images acquired every 2.4 seconds. But remember WATB. What are the blobs? To answer this, I'm going to explain three vital terms. Neurovvascular coupling, hemodynamic response, and bold signal. The key cells in your brain are neurons. These are neurons. They send signals and control your body. You have lots of blood flowing around your brain to provide neurons with oxygen. Importantly, when a group of neurons becomes very active and needs to send lots of signals, maybe like right now watching this video hopefully, the blood vessels nearby those neurons actually increase in size so that more blood, therefore more oxygen can be delivered to those neurons to meet their oxygen demands.

The activity of the neurons is coupled to the quantity of blood flowing to that area of the brain. This is called neurovvascular coupling. So when a person starts doing a task, because of neurovvascular coupling, the parts of the brain that begin to work harder to do this task also have the blood vessels in that area increase in size so more oxygenated blood can flow to that area. This is called hemodynamic response.

And we can actually plot this on a graph that looks like this. a delayed surge in blood supply as blood vessels get bigger and more blood flows to that part of the brain. Then a decrease as the additional required oxygen gets used up and then a slight undershoot. This process occurs across 4 to 6 seconds following brain activity beginning. So it is a delayed response. This process, this graph is called the bold response or bold signal. the blood's oxygen level dependent signal because this is a signal that depends on the level of oxygen in the blood that results from the different levels of activity in different parts of the brain.

So we have neurovvascular coupling where neuron activity is coupled to the quantity of oxygenated blood they receive. We have hemodynamic response where an increase in activity causes an increase in the size of the blood vessels. And we have the bold signal which measures changes in the magnetic field caused by changing levels of blood with and without oxygen. Now we're ready to understand the next stage. We have our fMRI images acquired every couple of seconds and we have the bold response going on inside the brain which means whilst the person is doing a task like tapping their fingers, blood with oxygen will flow to the parts of the brain that become more active and we're taking an fMRI scan of their brain every couple of seconds whilst blood with oxygen flowing to different parts of the brain is changing based on the amount of activity there.

Gee, if only blood with oxygen and blood without oxygen had slightly different levels of magnetism that could be detected by our giant brain scanning magnet. Oh wait, that is exactly what's happening. Blood without oxygen called deoxxygenated blood is what we call paramagnetic and distorts the magnetic field. Blood with oxygen called oxygenated blood is diamagnetic and does not distort the magnetic field. This is what allows us to track where more or less oxygenated and deoxxygenated blood is going in the brain whilst the person is doing a task.

This is actually what fMRI is measuring. And now the final stage to truly understand what the blobs are. We need to design a task or as the pros call it task design. We can design any task and this is where there's a huge amount of creativity in designing experiments. We could get the person to look at different faces, listen to music, spell words, tap their fingers, do maths, or get a person with a neurological disease like epilepsy or Alzheimer's and get them to do a task.

Let's use the example of fingertapping. For simplicity, we plot something like this, which is called a box car design that shows when the participant is resting and when they're tapping their fingers, usually for a period of 1 minute each, though this can vary. So this task starts with rest for a minute, then they tap their fingers for a minute, then they rest for a minute, then they tap their fingers for a minute several times over.

Whilst they do this task, we take fMRI images of their brain every 2.4 seconds. Now we can take the task design which is what they were actually doing periods of one minute rest followed by one minute of the task several times over and we combine the beginning of every task period with the bold response which sounds a bit complex but it's literally one line of code in Python. The result of combining them or convolving them produces this which is termed the canonical model.

But from here I'm just going to call it the model. This model in theory should reflect the exact activity of an involved area of the brain whilst doing the task. An increase in activity with the start of each task that's slightly delayed because of the bold response. The time for more blood with more oxygen to flow to that area of the brain is slightly delayed. Activity of an involved area of the brain in tapping should be doing this. high when fingers are tapping and low when fingers aren't tapping with a slight delay because of that bulb response for the oxygen to flow there.

And remember, we're collecting fMRI scans every 2.4 seconds whilst a person taps their fingers. fMRI scans which are sensitive to where blood with and without oxygen is traveling inside the brain. Our fMRI scan gives us a 3D whole brain volume every 2.4 seconds across time. So really 4D data. Every voxil or cube in the brain will have its own unique signal that can be tracked across time whilst the person does their fingertapping task.

So let's look at some real data. You can see in red our model that we built before which is exactly what a voxil or area of that brain should be doing if involved in the task. And in blue we have an actual voxil in the brain whilst doing that task captured by the fMRI scan every 2.4 seconds. If the two signals are very similar then it means the part of the brain where the voxil is is likely to be involved in the task.

So what do you think looking at this data? Do you think this particular voxil is involved in fingertapping? Yes, it absolutely is. We can now take every voxil in the brain and compare it to our model using some statistical analyses. If the two signals match really well, then we can give those voxels a certain color, let's say yellow. If they match medium well, then we can give them another color, let's say red. And if they don't match at all, let's just give them a gray color.

Do you see where this is going? Boom. This is what the blobs are. They are a reflection of whether those particular voxels match the model we created really well or really badly. The voxels which match it really well, we can infer are the brain areas very likely to be involved in the task. In this case, fingertapping. This helps us reveal through clever task design which parts of the brain are responsible for carrying out which tasks.

Just by the way, this blob here is the left motor cortex. The real brain activation given in a participant who was fingertapping with their right hand, which is exactly what we would expect as the left motor cortex is what becomes highly active to control right-handed fingertapping. What I've described in this video is something called a general linear model or a GLM, which is the most common way for analyzing fMRI data.

This is usually done in a language called mat lab in a library called SPM written by the most influential neuroscientist currently alive fristen. GLM can also be done in Python or I guess any language if you're smart enough. Uh this video doesn't contain the practical steps to actually carry out a GLM in SPM. But YouTube a moment here. If this gets 10,000 likes, it probably never will. Um, I'll make a video showing you how to build a GLM with fMRI data in SPM.

Also, the technology behind the blobs is most commonly fMRI, but it is worth noting that the methods I've described in this video can be applied with other brain scanning technologies like EEG or PET. Making these videos takes a lot of time. This one was about 75 hours. And if you want to support this content, please consider subscribing as it helps the channel out loads. You can also buy me a one-off coffee or become an ongoing supporter by joining my Patreon.

Links for both below. Thank you so much to the very generous people who have bought me a coffee and thank you to my amazing Patreons. Thank you so much for watching. Let me know if you have any questions. I can reply to all comments because this channel's small. And I'll see you in the next one. [music] >> [music]