tag:blogger.com,1999:blog-4402160631955197288.post6929032763406233618..comments2017-12-10T13:40:06.637-08:00Comments on practiCal fMRI: the nuts & bolts: Physics for understanding fMRI artifacts: Part ElevenpractiCal fMRIhttp://www.blogger.com/profile/07387300671699742416noreply@blogger.comBlogger16125tag:blogger.com,1999:blog-4402160631955197288.post-15434407150577284242017-11-19T14:37:42.593-08:002017-11-19T14:37:42.593-08:00Hi Luca, yes, it is possible to change the FOV and...Hi Luca, yes, it is possible to change the FOV and the matrix independently. There are two relationships at work simultaneously, given in the first two blue boxes. The first defines the FOV, the second defines the pixel size in each dimension, as FOV/N where N is the number of pixels in that dimension. So this is the part of the post where pixel size is introduced:<br /><br />Defining resolution is straightforward now that we have already got the FOV relationships established. All we need do is divide the FOV in each dimension by the number of pixels defining that dimension to see the k-space relationship:<br /><br />But I didn't call the value FOV/N the pixel size. If you re-read the post but always think of FOV/N as the pixel size, does it now make sense? You can see from the blue box defining FOV/N that the FOV and the pixel size are independent parameters, and keeping pixel size constant if FOV changes simply means that N - one dimension of the matrix size - must also change in concert.practiCal fMRIhttps://www.blogger.com/profile/07387300671699742416noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-16556391306418939632017-11-19T14:20:26.726-08:002017-11-19T14:20:26.726-08:00Wonderful blog... I learn a lot from that.
There i...Wonderful blog... I learn a lot from that.<br />There is a thing that I can't understand.. On the scanner is possible to change only the fov (increase/decrease pixel size) leaving matrix unchanged ?<br />reading the formula : pixel size = FOV / matrix I think it's possible <br />but when I read that : FOV = 1 / delta K in the image there is the same pixel size. I have change only the fov without changing pixel size. How is possible ?<br />Luca Antoniolihttps://www.blogger.com/profile/09385993325802063043noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-73219033583837209532016-04-29T01:27:30.193-07:002016-04-29T01:27:30.193-07:00Sure! Be glad to take a look. It's practicalfm...Sure! Be glad to take a look. It's practicalfmri at gmail dot com. I'm traveling thru 6th May but will take a look as soon as I'm able. practiCal fMRIhttps://www.blogger.com/profile/07387300671699742416noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-90508066984232703392016-04-28T17:50:42.917-07:002016-04-28T17:50:42.917-07:00Your blog rocks!
I've been puzzling for quite ...Your blog rocks!<br />I've been puzzling for quite a while about an image artefact that intermittenly appears in some of the DW images. Is there any way that I can send you the images for your comments? Thanks in advance.<br />limpeterhttps://www.blogger.com/profile/12170062222508855970noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-29622935305408772642012-08-17T08:46:34.120-07:002012-08-17T08:46:34.120-07:00Thank you again, the discussion was very useful fo...Thank you again, the discussion was very useful for me, could we please continue it by email? Mine is fabrybo@hotmail.comf4brynoreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-52797360327068114942012-07-06T15:38:07.366-07:002012-07-06T15:38:07.366-07:00"Multi-shot sequences decrease also the geome..."Multi-shot sequences decrease also the geometric distortions, as you mentioned in one of your post. Also in this case, there is a relationship between the geometric distortions and the number of shots?"<br /><br />In case I haven't been clear, the critical parameters for distortion extent are the inter-echo time, i.e. the time between each line of kx sampling, and the size of the step in ky between each echo. For single-shot EPI, each new echo in the train defines a step in ky (the phase encode dimension) of size delta-ky.<br /><br />Any change that decreases the inter-echo time or increases the ky step size between each echo will tend to decrease the distortion, all other parameters held constant. For example, we could simply skip every other ky step, i.e. make the step size R*delta-ky, as is done in parallel imaging methods such as GRAPPA for R-fold acceleration. (The amount of phase evolution between ky samples has been reduced by a factor R, thereby reducing the distortion by a factor R.) Likewise, if one does an interleaved multi-shot EPI scheme then one can achieve similar n-fold reduction of distortion for an n-shot acquisition, but only if the inter-echo time is maintained and the ky step size is increased by the factor n, to n*delta-ky. (The interleaved k-space then has to be combined prior to 2D FT.)<br /><br />Other "go faster" EPI methods, such as partial Fourier EPI (post to come), don't decrease the distortion extent in the phase encoding dimension because the ky step size and the inter-echo time are unchanged. And it's also possible to acquire some segmented multi-shot EPI schemes where the ky step is unchanged, and those don't decrease distortion either (although they can permit higher resolution than single-shot EPI). <br /><br />I'll add a post on distortion and methods to reduce it to my list. There's too much information to portray here!practiCal fMRIhttps://www.blogger.com/profile/07387300671699742416noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-88027618217570784472012-07-06T12:50:29.328-07:002012-07-06T12:50:29.328-07:00"When you say "inter-echo time" do ..."When you say "inter-echo time" do you mean the time to move from one kx-line to another?"<br /><br />Yep!<br /><br />"And when you say that each line is acquired every 2ms...do you mean a shot trajectory every 2ms?"<br /><br />Correct, it would be one entire line of kx (read axis) in my example.<br /><br />"As you said in your post the image resolution is FOV/N = 1/(N Dk) = 1/kmax. If we keep the FOV fixed, we should increase the number of points acquired to increase the resolution, shouldn't we?"<br /><br />Yes, and to do that we need to push out farther into higher values of k-space.<br /><br />"Could we say that with more shots we have more points in the k-space?So, if n are the number of shots, FOV/N=1/n(N Dk) ?"<br /><br />Ultimately the number of points in k-space is the same whether those values are acquired in one go - single-shot EPI - or via some sort of multi-shot scheme. Also, the actual k-space values acquired in the final 2D k-space matrix are the same regardless of the number of shots. We're simply chopping up the acquisition into smaller chunks in order to attain some experimental gain. There aren't many good articles on multi-shot EPI that I've found, but Stuart Clare's PhD thesis might make a useful next step:<br /><br />http://users.fmrib.ox.ac.uk/~stuart/thesis/chapter_5/section5_2.html<br /><br /><br />"A multi-shot EPI is not longer than a single-shot EPI?"<br /><br />Yes, multi-shot EPI takes n times longer for n shots, which is a big reason why multi-shot EPI isn't common for fMRI.<br /><br />"Multi-shot sequences decrease also the geometric distortions, as you mentioned in one of your post. Also in this case, there is a relationship between the geometric distortions and the number of shots?"<br /><br />Yes, all other parameters being equal the geometric distortion extent will be inversely proportional to n. But because of the increased time to acquire each final image, and the increased motion sensitivity that extending the total imaging time takes, it's also not a common tactic for fMRI.practiCal fMRIhttps://www.blogger.com/profile/07387300671699742416noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-67381540683839889472012-07-06T05:55:03.426-07:002012-07-06T05:55:03.426-07:00Great! I will read it willingly!
thanks,
f4bryGreat! I will read it willingly!<br />thanks,<br />f4bryf4brynoreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-3588949576272153302012-07-06T05:52:48.838-07:002012-07-06T05:52:48.838-07:00Thanks a lot! It is very useful!
I understood the...Thanks a lot! It is very useful!<br /><br />I understood the idea but I have some problem with the nomenclature and to "see" that in the k-space. <br /><br />When you say "inter-echo time" do you mean the time to move from one kx-line to another? And when you say that each line is acquired every 2ms...do you mean a shot trajectory every 2ms?<br /><br />As you said in your post the image resolution is FOV/N = 1/(N Dk) = 1/kmax. If we keep the FOV fixed, we should increase the number of points acquired to increase the resolution, shouldn't we?Could we say that with more shots we have more points in the k-space?So, if n are the number of shots, FOV/N=1/n(N Dk) ?<br /><br />A multi-shot EPI is not longer than a single-shot EPI?<br /><br />I'm sorry to bore you with all these questions, but I have another question ;)<br />Multi-shot sequences decrease also the geometric distortions, as you mentioned in one of your post. Also in this case, there is a relationship between the geometric distortions and the number of shots?<br /><br />Thanks you very much for all your answers!<br />f4bryf4brynoreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-37345579003772977362012-07-06T04:28:47.984-07:002012-07-06T04:28:47.984-07:00Thank you very much, I've got it!
f4bryThank you very much, I've got it!<br />f4bryf4brynoreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-91081209099396912832012-07-05T11:10:32.256-07:002012-07-05T11:10:32.256-07:00f4bry, a post script: One oft overlooked point abo...f4bry, a post script: One oft overlooked point about resolution is the intrinsic blurring of pixels by T2* decay under the EPI acquisition train. This is the so-called "point spread." I'll be doing a separate post on point spread at some point. But it's useful to remember that the nominal resolution, defined as FOV divided by the number of pixels, is always narrower than the effective resolution. Pixels are always smoothed somewhat by virtue of the physics of the acquisition.practiCal fMRIhttps://www.blogger.com/profile/07387300671699742416noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-48268780126986633532012-07-05T11:07:12.712-07:002012-07-05T11:07:12.712-07:00f4bry, on your second question about multi-shot (i...f4bry, on your second question about multi-shot (interleaved) EPI, the root of the reason is the duration of usable signal after an excitation RF pulse. The typical T2* of brain tissue is in the range 20-60 ms at 3 T. Thus, after a best case of some 100-150 ms the signal level is rapidly approaching zero, and an increasing level of noise would be detected. <br /><br />Now think about wanting to acquire an EPI that is, say 256x256 pixel with a field-of-view of 224 mm. That is equivalent to a nominal pixel resolution of 0.875 mm. Now, to acquire each line of kx-space (the read direction) will take four times longer for 256 readout points as for 64 readout points. If 64 points causes an inter-echo time of 0.5 ms then each line of kx-space takes 2 ms to acquire. We might be able to acquire something like fifty total echoes before we simply run out of signal. We can't hope to acquire all 256 phase-encoded echoes in a single shot! The signal-to-noise would be awful, tending towards unity! <br /><br />But if we chop up the 256 phase-encoded echoes into four groups of 64 echoes each, and combine the four shots before doing the 2D FT, then we have a fighting chance of producing a final image with appreciable SNR at the target nominal resolution. Overall, then, it's a signal (or T2*) limiting situation. If there were a way to make T2* longer then, in principle, we could acquire higher spatial resolution in a single shot. Sadly, however, we're inherently limited in our ability to shim the brain and prolong the T2*.<br /><br />Cheers!practiCal fMRIhttps://www.blogger.com/profile/07387300671699742416noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-36497486987647709852012-07-05T11:00:09.675-07:002012-07-05T11:00:09.675-07:00Hi f4bry,
On your first question, I think I se...Hi f4bry,<br /><br /> On your first question, I think I see the source of your confusion. The crucial phrase is "either side of zero." So we do actually acquire Nx k-space points for Nx points in image space; Nx/2 of them are on the negative side of kx-space, Nx/2 of them are on the positive side of kx-space (if you ignore for simplicity the usual convention of acquiring one point at kx=0, which results in one fewer k-space point acquired on one side of kx when Nx is even). <br /><br /> On another tack, be careful trying to associate each point in k-space with a single point in image space. In fact, one point in k-space represents every point in image space that contains the particular spatial properties of that point in k-space. That's why it's useful to demonstrate what happens in an image when certain regions of k-space are set to zero, as in the first figure in the post. If just one point of k-space were set to zero it would have a (potentially) universal effect on the image; "potentially" because only if that particular k-space point represents spatial frequencies present in the image would there be a non-zero value at that k-space point.practiCal fMRIhttps://www.blogger.com/profile/07387300671699742416noreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-34198717535592585882012-07-05T09:15:14.763-07:002012-07-05T09:15:14.763-07:00I have another question : why is it possible to ob...I have another question : why is it possible to obtain higher spatial resolution with a multi-shot (interleaved?) EPI? <br /><br />Thank you,<br />f4bryf4brynoreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-52899790854807852202012-07-05T04:58:55.848-07:002012-07-05T04:58:55.848-07:00Yes, great post!
I didn't understand this sent...Yes, great post!<br />I didn't understand this sentence "Remember that for an Nx x Ny image we acquire Nx/2 and Ny/2 values of k-space either side of zero, making the total span of k-space equal to 2kxmax and 2kymax."<br /><br />Maybe it is obvious, I'm sorry. Why we acquire Nx/2 points in the k-space for Nx points in image domain?Each point of the real object corresponds to a particular point in the k-space, isn't it? Can you try to explain me that in other words?<br />Thank you in advance,<br />f4bryf4brynoreply@blogger.comtag:blogger.com,1999:blog-4402160631955197288.post-74998265296258200122012-04-27T11:45:16.791-07:002012-04-27T11:45:16.791-07:00great post. really helped me a lot. keep up the go...great post. really helped me a lot. keep up the good work...Anonymousnoreply@blogger.com