版 版

版权声明：转载时请以超链接形式标明文章原始出处和作者信息及本声明
http://banban.blogbus.com/logs/31837716.html

Q: Hi Cisy! I'm from your tutorial group and would like to ask you a question regarding fMRI. I understand that fMRI is said to be a good method as it provides structural and functional information about the brain. But how are such images reflected on the image? Is it right to say that let's say the subject is doing some task (e.g. reading) then the fMRI image will show functional information as it reflects the areas activated?

 A:  Hello ***, the basic idea in the fMRI data analysis is subtraction. Usually in the experiment design, there are two types of conditions: experimental condition and control condition. We measure BOLD signals from the brain during the control condition and the experimental condition. Then we subtract the signal values in the control condition from the signal values in the experimental condition. The signal difference between the two conditions may be significantly bigger than zero in some brain regions (e.g., region A) but not the others (e.g., region B). We attribute this signal difference to the difference between the condition manipulation (e.g., viewing face pictures vs viewing random dots pictures). And the related brain region is A, not B.

The brain images we see in fMRI papers are actually a reconstructed image from the BOLD signals. Imagine we divide the brain into many many 2-by-2-by-2 mm^3 cubicles and we call them voxels. From every voxel, we can obtain a signal value, which varies depending on the materials (e.g., white matter or grey matter), the activation (e.g., excitation or inhibition), etc. Based on these voxel values, we will be able to know the types of materials in a specific region. If we measure the signals every 2 seconds, we will be able to know how the activation changes over time.

The minimum time interval between two measurements depends on the strength of the magnetic field, usually 2 or 3 seconds in a 3 Tesla magnetic field. That's why we say fMRI doesn't have good temporal resolution. To increase the temporal resolution, we need an fMRI system that can offer stronger magnetic field (now the newest model can offer 7 Tesla, but super expensive). But since the voxel size can be around 2-by-2-by-2 mm^3, we say this is quite a good spatial resolution.

Q:  I understand the part where you mentioned about the temporal and spatial resolution. But my question is what is the difference between structural and functional information produced from fMRI. Or is there no difference between the two?

A:  Well, this is really a long story. I'll try to cut it short below.

In terms of what the fMRI is looking at, they are different. Both the structure and the function scans are related to how the protons in the brain respond in a magnetic field, e.g., atom of hydrogen with a single proton, as our brain is in quite a liquid environment. By sending pulse to set the protons into high energy state (excitation), we observe how the protons respond and return to the low energy state.

There are various parameters we can measure. For example, the rate at which the proton recovers from the high energy state to the low energy state after excitation, called T1. The structure scan usually looks at T1, e.g., cerebrospinal fluid has a T1 value of 4000ms at magnetic strength of 1.5T, while white matter with T1 of 600ms at 1.5T. Also, under normal conditions, thermal energy causes the proton to spin about itself, but the protons all have different phases and random directions. After excitation, the protons will have the same phase at the beginning and as they lose energy they become out of phase. The rate of this loss of phase coherence is called T2. The speed of losing phase coherence is affected by the interaction between protons, related to what the substance is. Thus, in the structure scan, we also look at T2. Both T1 and T2 are considered constant rates for a given substance. However, the speed of losing phase coherence can also be affected by the local magnetic environment surrounding the substance. So the combined effect of both interaction between protons and the local magnetic inhomogeneities are described by the T2*. The function scan usually looks at T2*.

We talks about oxy-hemoglobin and deoxy-hemoglobin in fMRI experiments. The oxy-hemoglobin is diamagnetic, which does not affect much the strength of the surrounding magnetic fields. However, the deoxy-hemoglobin is paramagnetic, which can distort the local magnetic fields. As we know, neural activities will affect both the amounts and the proportions of the oxy-hemoglobin and the deoxy-hemoglobin. So local neural activities will affect T2* observed in that area.

Well, it is really a long story. To avoid turning this email into a book, I omitted a lot of details such as how to determine the time interval of sending the pulses, how to localize the signals, etc., which link too much to the signal processing theories. If you are interested in these topics, books are really helpful (e.g., Jezzard, Mathews & Smith (2003), functional MRI: an introduction to methods).