Introducing MRI: Phase Encoding and k-space (23 of 56)

www.einstein.yu.edu - The twenty-third chapter of Dr. Michael Lipton's MRI course covers Phase Encoding and k-space. Dr. Lipton is associate professor radiology at Albert Einstein College of Medicine and associate director of its Gruss Magnetic Resonance Research Center.

[@benyamindeldar9011]

i think im slightly in love with this man. THANK YOU!

[@841533356v]

benyamin deldar 😄

[@ya_boy_jarvis8307]

The only thing more amazing than the brilliant minds that made this machine is the fact that this guy can explain it in a way I can understand it. Thanks professor!

[@palmtreebeautiful]

Dr. Lipton is really a great teacher. Can you please upload the whole series? Thanks!!

[@Belliisan]

Now I finally understand what K-space and the Fourier transformations are, thanks for uploading it! :D

[@paulokot1235]

I don't know why you are not receiving the like you deserve.

[@keyaamarsee9631]

Thank you so much for these videos. They help so much.

[@nixgeldring3564]

Thanks for the great series. Right at the level I need it :). I have 2 questions: 1. In the beginning, one of the students proposed taking 2 measurements, each one with the frequency encoding in a different dimension (e.g. x and y). Dr. Lipton mentions that this has been done, but phase coding is usually used nowadays. Phase coding involves taking many different measurements (way more than 2) with slightly differing phases, and the resolution is limited by the number of measurements taken (while the resolution along the frequency encoded axis is only limited by how fast we can sample). My question is, how is phase encoding more efficient? Or if efficiency isn't the goal, what is the goal? 2. I thought K-space was a plot of the fourier-transformed space, but here the first plot is shown with time as the x axis. I don't understand how k-space can be symmetric if we are in the time domain. The signal is still decaying, so even if there is a bump due to the re-phasing of the signal, it can't be exactly symmetric. I think one of the students asked this, and I didn't really get the explanation. .... other than those few lingering questions, a very useful lecture!

[@rubenvandenbroek9689]

not entirely sure, but I can try to answer this: 1 machines have been getting better and better, being able to do many phase encoding steps during one TR.

[@giuseppeturco7377]

Sorry, regarding the graph at 15:23, there would be the magnitude of the magnetic field on the ordinate axis, not a gradient, right? A gradient is expressed by a variation on the magnitude of magnetic field depending on the position in consideration; by plotting a variation of a gradient in function of a position, you are expressing a "gradient of a gradient" of a magnetic field. Am I wrong? Furtherly, what do you mean with "isocenter"?

[@vijaymittal7998]

thanks for uploading, great clarification

[@EricDiazMD]

I'm not sure why you say that the "dimension" of the x-axis in K-space (aka., Kx) is time and that the center of Kx is equal to TE (@23:50 and @42:03). You also mention this in the next video (@20:10). I was under the understanding that K-space is in the frequency domain, essentially AFTER you do the Fourier transform of the time domain signal.

[@practiCalfMRI]

K-space is the time domain, it is the image that is in the frequency domain. But here's the confusion, perhaps. The image is labeled in cm, not frequency. So the conjugate variable for the image is in 1/cm, which is the units of k-space. The units of k-space, 1/cm, have a time domain equivalence through a simple unit conversion. More details here: practicalfmri.blogspot.com/2011/07/physics-for-understanding-fmri_29.html

[@salvatore4219]

I literally love your lessons! Just one thing! Personally I think the fact that is necessary to turn on twice the encoding phase gradient for spatial localization should have been explained a few minutes before! When I finally heard, it was like "now it's clear!". Anyway...thank you so much for your lessons!

[@roycolton4537]

the phase encoding gradient is turned on with every new iteration of pulse sequence, but with a corresponding change in its strength. I think you are mixing up. Dr Lipton demonstrated the phase change after two successive applications of the phase-encoding gradient (from two different iterations of the pulse sequence), to make sure that it's the *change* in phase, not the phase itself, is what is essential for the Fourier transform.

[@rfmonkey4942]

absolutely brilliant !

[@vmjable]

@ 30:00 my head literally exploded. Im confused. If you draw a line at the iso-center along the y -axis and symmetrically move outward toward the top and bottom of the slice by +1 & -1 row....are those corresponding rows that are equidistant from the isocenter experiencing the same magnitude/difference of phase incoherence? I hope my question was not too confusing. Please help. Thanks

[@fatima.aslamm]

No there will be difference bcz as we move from positive side of the gradient magnetic field to negative side there is decrease in the net magnetic field. magnetic filed magnitude will not be the same at both sides of isocenter

[@zakabhatti7243]

@@fatima.aslamm I am a lil more confused about isocentre(s). Do we have two of these (x & y)?

[@dangerousman4071]

the direction of phase incoherence will be different, even if the magnitude will be same when you compare two equidistant sides from the isocenter, remember that it is a vector. or you could say that the magnetic field strength is decreasing linearly and the isocenter is where the field strength is same as Bo. So it will different at different points on this linear scale. either way to think about it is ok.

[@aketisrinu3541]

I have one doubt. When we turn on and off phase gradient there will be a phase shift changing from bottom to top. Is this change in phase shift (in the signal) not enough to find the local (spatial) information?. Why we need to apply different phase gradients. Is phase gradient increase gradually from bottom to top or not? If yes, is phase shift changing gradually from bottom to top or not? If yes, why the difference in phase shift is differ between two points which are near and far from Isocenter?. why don't the difference be same? Kindly help me in this regard. Thanks in advance

[@mandolinic]

An important rule of the Fourier transform is that does it does not generate new data. You have to put into it as much data as you want to get out. So, if you want to break a signal down into 256 separate frequencies using a 1D DFT, you need to supply 256 pieces of data. Expanding this into 2 dimensions, if you want to create an image which is 256x256 pixels then you need to start with a grid of data that is 256x256. A single read of signal data (even with an applied phase encoding gradient) is only going to give you a grid of data which is 256x1. Well short of the 256x256 data values you need to build a 2D image. Hence the need to make one read for every row of pixels in the final image. The really clever bit is that someone noticed that shifting the phase gradient and using a 2D DFT gives you a displayable image. I'm reasonably familiar with image processing and DFTs, but I would never have thought of that!

[@practiCalfMRI]

Try these posts to see if it clarifies things for you: practicalfmri.blogspot.com/2011/07/physics-for-understanding-fmri_29.html practicalfmri.blogspot.com/2011/08/physics-for-understanding-fmri.html practicalfmri.blogspot.com/2011/08/physics-for-understanding-fmri_15.html I've taught psychologists the basis of fMRI for nearly two decades, and I've learned that some non-traditional approaches to k-space can help. In particular, coming at it backwards, that is, start with a 2D image and see how that appears in k-space. Doesn't even need to be MRI.

[@practiCalfMRI]

"The really clever bit is that someone noticed that shifting the phase gradient and using a 2D DFT gives you a displayable image. I'm reasonably familiar with image processing and DFTs, but I would never have thought of that!" Seconded! What is even more amazing to me is the fact that the way of doing phase encoding - what was initially called spin warp imaging - came about before the k-space description of MRI came about! Indeed, all the commonly used forms of spatial encoding came about before the k-space pictorial representation we all use today. That makes the invention of these spatial encoding methods all the more remarkable. If there was no k-space representation, understanding this stuff would be mind-bendingly hard. (As I found out when trying to understand EPI from the original Mansfield articles. Don't try this at home! Use k-space. It is your friend!)

[@natjimoEU]

great teaching, very smart guy.

[@dangerousman4071]

how is it that 256 iterations are required for phase info if we need only difference in the dephasing caused by different strengths of G. Wouldn't just two iterations be enough? to get the difference in dephasing along the phase encoding direction? and that difference would obviously be a gradient with respect to the isocenter. why 256 iterations for 256 voxels along phase encoding direction? plz someone explain.

[@as180697]

Amazing!

[@kamalhassan8358]

Example of radio frequency in mri?

[@jenisundaram]

Superb

[@jhk312]

thank you !

[@mohmad801]

nice lecture how to upload

[@mohha1820]

einstein of the mri dr michael

[@user-zb8nw1py6k]

It is good lecture

[@whytemaamba]

he's gotta be making this s#it up as he goes along.

[@balioxxo_8426]

You lost me doctor

[@jacobvandijk6525]

Why is he talking about the amplitude of a gradient??? Just call it its magnitude. Amplitude is the amount of signal being received. Leave it that way. Don't confuse people by introducing unnecessary and confusing "amplitudes".