Measuring SNR with B1 and P
I am trying to simulate a receive coil for MRI and I am stuck in measuring SNR. I came to know that SNR can be found by dividing B1 field to dissipated power. But, the unit of B1 field is Tesla and dissipated power is watt.
I know that as long as we do a relative comparison, we can calculate SNR. But, I am really confused at a simple matter. Should I be taking the ratio of Tesla to watt ?
Or I should be converting to microtesla and kilowatt before taking the ratio.
Any kind of suggestion is really appreciated.
Thanks
It does not work that way. The signal depends on many factors of proton emission from the spin relaxation time and density of volume pixel or voxel, field of view and many other factors. The magnetic field Bo in Tesla is longitudinal but the detecting signal is tangentially deflected. The noise level of resistance in the coils, amplifier and background are required to start looking at SNR. But there are many enhancements like phase modulated bursts to echo cancel fat tissue response. Then there are sampling interval variables, slice pitch, statistical averaging, patient dose levels, Signal bandwidth, motion artifacts etc that each contribute to SNR.
Here is a basic primer . http://wiki.ci.uchicago.edu/pub/HNL/...oesmriwork.pdf
Thanks, but what if i want to take the ratio. What value should i use. Microtesla and kilowatt or microtesla and watt.
1T=1Vs/(m^2)
1W=1VA
So the ratio's unit will be s/(A*m^2)
SNR of an NMR experiment is a very complicated thing, and there are various ways of defining it (image SNR, k space dynamic range, etc). It's not an inherent property of any coil, and will of course depend on the phantom properties.
What are you using as a basis for SNR calculation? Where are you evaluating B1? You're looking at B1-, not just B1, correct? Are you referring to power dissipation in the coil or the phantom?
I am using Remcom software for calculating SNR. I am referring power dissipation and B1 field with respect to phantom. Do you think it is possible to evaluate in such a way?
Also, if you know any other way through which I can visualise SNR, it would be really helpful.
Well first you can't define the absolute SNR of a coil, because SNR is a property of a signal, not a piece of hardware. What you can measure are figures of merit for the coil/phantom setup which will be proportional to the SNR. One of the simplest FOMs is B1 power efficiency, in which you inject current into the coil to transmit a B1 into the phantom and measure the B1+ and B1- (as functions of space) and the total power absorbed at the coil terminals (in watts). The ratio of B1 to the square root of power is a figure of merit for coil sensitivity, and thus SNR. Using B1- gives receive sensitivity, using B1+ gives transmit sensitivity. It's still not a measurement of SNR, but it is a common figure of merit for comparing the sensitivity of coils.
Thank you for your reply again.
I guess the unit of B1 is not essential in this case. It can be used as microtesla or tesla as it only shows the proportionality, right ?
So, is it okay, if we use B1 in microtesla and power in watts to take the ratio for comparing the sensitivity of coils?
Yes, but you use the formula B1/√(P) not B1/P (since power should always be proportional to B12). Note that this is only valid for comparing coils when the phantom is exactly the same.
Thank you.
What if I want to visualize the SNR?
I think that I will need to post-process the data in Matlab. Do you know where I can get the code for post-processing the data so that SNR can be visualized or from where I can get started?
First you have to define SNR. Are you talking about the SNR of an acquired image? The dynamic range of a received signal? The sensitivity of the coil?
I am trying to visualize the SNR for the sensitivity of the receive coil.
SNR and coil sensitivity are different things. SNR isn't even arguably the best way to evaluate coil sensitivity, even in a receive coil. If you want, you could set up an experiment with a phantom that generates uniform magnetization, apply an imaging sequence, simulate the received signal, reconstruct the image, then use that image as a sensitivity map. But it's a lot simpler to just use the approach described above.