In the first few minutes of this video I take some time to recap what we've covered so far. For those looking to get straight to phase encoding, skip to 5:56
I have never seen any yt channel explain that complicated physics that well, it is so precise but so clear in the same time, thank you very much for your work !
I have never commented on a TH-cam video before but I wanted to thank you so much for making these concepts so much easier to understand. I am better off watching your youtube videos then my lectures from my professors. Keep up the fantastic work. I am sure I am not the only one who would like to give you credit.
That is the best toutorial series i have ever seen on TH-cam (and inwatched many about many topics). So detailed, but not in a typical lecture style from university. Thank you so much 🙏
Like all the comments I've read for this series, I'm so appreciative of the clear way that you explain these complicated concepts and have shared this series with my radiation oncology colleagues that are looking for MRI refreshers as they are much easier to understand than any lectures or course that I have ever attended. Well done and thank you!
Thanks for these amazing videos! I'm a radiology resident here in the US and I literally have made some (correct) calls I wouldn't have because of your videos!!
One thing I picked up. When at the beginning you say that in the absence of any gradients or RF pulses the spins are rotating in phase, that is not true. It is the application of 90 degree pulse that flips them and puts them in phase. Correct me if wrong but that is my understanding as a physicist.
Thank you. Each video takes me in excess of 10-20 working hours to create. I make all the illustrations by hand, need to research and plan the best way to deliver the information, need to film, edit, colour grade, render, upload and render on TH-cam, create a thumbnail etc. There’s just simply no way I can do daily uploads 😅
@@radiologytutorialsI know personally how long this stuff takes to make (a 3-5 min video for me takes about 6 hours 😂) and am in awe at the hard work you’re putting in. You’re doing great things here. Keep going Michael!
Thank you! The video is very helpful, I just don't understand why you say that in order to convert the time domain into frequency domain we use an inverse FT instead of a FT. Thank you again! If anyone else knows the answer, that would also be helpful :)
I have a question, could you please consider answering. At 6.39, you said the spins are spinning in-phase with one another, since the only magnetic field they experience is the main magnetic field. In the very first video you said and also showed with an animation that, with the influence of the main magnetic field, most of the spins will align parallel to the main magnetic field. They will precess at the same frequency (Larmor frequency), but not in-phase with one another.
I have the same doubt. if the spins thend to dephase the animation in sincrony confuses me so why the gradientes if each voxel have tissue with diferent decaiyng.
Another question... it seems you overlay the frequency encoding chart with k-space when explaining how to get the final picture, is this accurate conceptually? Wouldn't this mean that a frequency encoding chart has a Y-axis? I thought that any frequency encoding basically generates rows all having the same frequency essentially just generating only more x-axis rows?
Thank you Dr. A question: It is written in Sarah that in repeat cycles of phase encoding gradient, each point along the segment will have DIFFRENT FREQUENCY. Those that are at furthest ends will have higher frequency phase encoding curve. Those frequencies will differ when we apply frequency encoding gradient? You mentioned that frequencies will be same. These will be same before applying frequency encoding gradient?
Great observation. There are two separate frequencies being talked about here. The frequency of precession along the x axis will remain the same for each acquisition. The frequencies mentioned in the phase direction refers to the 2D frequency waveform vectors formed by both the imaginary and real components of the spins in the entire slice - these are dependent on the combination and phase and frequency gradients. In this course I have largely omitted 2D waveforms in order to avoid confusion. In the k-space talk when I mention the rate of change of frequencies increasing at the periphery of k space, I am referring to these 2D phase frequencies. Hope this hasn't caused you more confusion 😅
Brilliant video thank you! One thing I didn't understand about the two diagrams around minute 30 (k space vs. after first Fourier transform): why does the diagram on the right not have rows with less and less intense signal at the top and the bottom? Those were acquired with a stronger gradient so I thought they'd have weaker signal?
At the 7th minute when you were explaining that the magnetization vectors within a slice were in phase with one another, I got confused. I thought all the vectors would be out of phase until they reach TE after the 180 degrees RF pulse, given that different tissues may exist within a selected slice?
Congrats great course Im almost there to get it. but please answer, why the gradientes are needed if the spins thend to dephase by themself by the t2 decaiyn .the video off the spins in sincrony confuses me. wont the dephasin itself would create a signal in y and x because each voxel have tissue with differente decaiyn rate?????
Hi there, I am trying to put together the bigger picture: the phase encoding gradient is applied to the entire slice, and spins in the phase direction change their rate of precession depending on their location embedding a phase shift, when we apply the 180 rephasing pulse at a certain point of relaxation, does it only rephase the spins that still have some transverse mag? How does the addition of the gradient change the echo if the spins all align anyway with rephasing? My understanding so far is that with a certain amount of relaxation, we get different amounts of signal depending on tissue, I that we have an echo that is a bunch of different signal strengths based on T1 and T2 differences, but how does the change of frequency along the phase encoding gradient fit into this?
Why do we have to take so many phase signals? Why can't we just do FED in x-axis in one cycle? and FED in y-axis in second cycle and this will give us y axis values.
Thank you for uploading these helpful tutorials! I am confused about why FEG changes the frequency and PEG changes the phase of the spins, considering they both induce gradient fields?
Great question. PEG also changes the frequency for the time the gradient is on. However, once the gradient is turned off the spins will continue to precess at larmor frequency but will be out of phase. The FEG changes frequency (and is applied whilst we are reading the signal - therefore that frequency change is registered in the signal). Hope that makes some sense 🤞🏼 let me know if you need more explanation 🙂
I get a little confused when you say something like "256 different pixels." To me, a "pixel" is a final representation of an image voxel in the matrix. Like a camera picture, it's the smallest form of the grey scale which makes the image. Do you mean it more like "256 data points" like a numerical value of signal intensity? As an MRI tech, when I change the frequency matrix, I'm thinking I'm changing the number of frequencies I want to sample. For example, if the default frequency encoding number is 256, I think that means I'm telling the scanner to use a gradient and signal math to give me 256 different frequencies across the slice. The data gathered in that gives the scanner 256 differences to measure across the slice. And if I want to half that, I can choose 128 and now I get 128 different frequencies across the slice. It's like I'm literally choosing a data matrix across the slice. Am I thinking of this correctly? Or am I missing something about how K space is filled from the frequency encoding? TIA!
To clarify, I think of a 256 frequency matrix as dividing the slice along the frequency direction into 256 "columns" of different frequencies. Half of that would be 128 "columns" across the same space, meaning less frequencies are used across the slice to encode the exact same space the 256 was. This would mean we are creating 128 detectable differences in the slice that we can then use to associate with the known applied gradient to give us spatial location along a single axis. Dividing this info into pixels doesn't quite make sense to me, and I'm hoping you can help me understand. I'm also a bit confused as to how k-space is filled and cross-referenced with the frequency data. I thought frequency data was included in k-space. I haven't reached your video on that yet, so the answer to my question might very well already be out there!
Sorry for the delay in replying. Your thinking is correct. Mostly these terms can be used interchangeably but there are subtle differences. Can't remember exactly how I phrased it in this talk (I may have caused some confusion). I try to use matrix size and pixels along the x axis as synonyms because both refer to the spatial resolution in the field of view. The matrix size determines the number of voxels and ultimately pixels we will display within the FOV. The data acquisition sample number (ie the number of digital data points we sample from the analogue signal) determines the number of frequencies we can separate using a Fourier transform. Some will call this the frequency matrix. You're right, the number of times we sample a signal will determine the number of different frequencies we can delineate and ultimately place on an x axis location. However, there are sequences where we oversample the frequency encoding direction and disregard data that falls outside of the field of view (like when we oversample to avoid aliasing). In this case the frequency matrix and the image matrix do not correspond. I may not have answered your question. If not, let's chat some more!
@@radiologytutorials I appreciate your reply , and thanks for telling me I'm kind of on the right track. I think what's hard about the pixel reference for me is that to encode 2 different locations within the same frequency, the phase encoding gradient has to come in to play to differentiate them. I think in the following video you clear all of this up pretty well. The hardest thing for my students to get over is this thought that we're still recording an image in MRI. I literally have started telling them "none of this is real!" lol. When they really grasp that there is no picture, that our images are graphical representations of electrical signals, I think it really helps them start understanding how chemical shift, aliasing, and other artifacts are just electrical signals the computer can't process correctly. Just today I was having to scan at the edge of the magnet bore, and the computer has a really tough time encoding signals at that extreme of a gradient. The fat saturation was abysmal, and there was signal cut off. The scanner ran frequency encoding pre-scans for like 2 minutes trying to figure it out, and even then there were issues. It's a challenge, but it sure isn't boring!
Whilst the phase encoding gradient is on it is essentially creating a difference in frequencies along the slice. However, when the gradient is switched off all the spins now precess at the same frequency because the only magnetic field is the main magnetic field. The spins now have different phase but the same frequency. This differs from the frequency encoding gradient because we apply that gradient whilst reading the signal - therefore these frequency differences will be in our data. Hope that helps 🙂
basically its the on and off of the gradients thats key. Putting it on does change the frequencies along the Y-axis AND keeps the phases and then the turning off stores the phase but the frequencies all realign as they are experiencing a uniform magnetic field. Did I get this right?@@radiologytutorials
@@johnnysilverhand1733 thats only due to the time for relaxation of hydrogen atoms and local magnetic field effects avoided by not scanning evetything at once. the time component is not fundamental to the signal. without those factors, if we made the phase encoding steps 500ms, 5000ms, or 1 hour apart it will look the same as long as nothing in that region changed significantly
In the first few minutes of this video I take some time to recap what we've covered so far. For those looking to get straight to phase encoding, skip to 5:56
Hello sir where is the link of the question?
I have never seen any yt channel explain that complicated physics that well, it is so precise but so clear in the same time, thank you very much for your work !
Thank you Lucie! That’s so kind of you 🙂
I have never commented on a TH-cam video before but I wanted to thank you so much for making these concepts so much easier to understand. I am better off watching your youtube videos then my lectures from my professors. Keep up the fantastic work. I am sure I am not the only one who would like to give you credit.
Hi @allyharrison4163 - thank you for such a lovely comment! Really helps to keep me motivated to keep going! Glad you've enjoyed them 🥳
for anyone confused,
upon watching again
it starts to make sense!
That is the best toutorial series i have ever seen on TH-cam (and inwatched many about many topics). So detailed, but not in a typical lecture style from university.
Thank you so much 🙏
Like all the comments I've read for this series, I'm so appreciative of the clear way that you explain these complicated concepts and have shared this series with my radiation oncology colleagues that are looking for MRI refreshers as they are much easier to understand than any lectures or course that I have ever attended. Well done and thank you!
This is brilliant. Thank you so so so much. I’m preparing my PhD qualifying exams and you’re being a lifesaver!!
Excellent! Glad they're helping 😊
Thanks for these amazing videos! I'm a radiology resident here in the US and I literally have made some (correct) calls I wouldn't have because of your videos!!
Wow! That's great to hear!
One thing I picked up. When at the beginning you say that in the absence of any gradients or RF pulses the spins are rotating in phase, that is not true. It is the application of 90 degree pulse that flips them and puts them in phase. Correct me if wrong but that is my understanding as a physicist.
You’re right. I must have misspoken! Phase coherence only occurs with excitation and resonance secondary to the RF pulse 👍🏼
Thankyou sir for best content ever on youtube
Dear Michael, you are doing a great job! Kindly upload at least one video lecture daily...regards
Thank you. Each video takes me in excess of 10-20 working hours to create. I make all the illustrations by hand, need to research and plan the best way to deliver the information, need to film, edit, colour grade, render, upload and render on TH-cam, create a thumbnail etc. There’s just simply no way I can do daily uploads 😅
@@radiologytutorialsI know personally how long this stuff takes to make (a 3-5 min video for me takes about 6 hours 😂) and am in awe at the hard work you’re putting in. You’re doing great things here. Keep going Michael!
Thank you Naveen! You’ve been a great inspiration for me 🙏🏻 Appreciate the encouragement 🙂
thank you so much for the great teaching, i wish you had teaching also on CT physics!!!!!!!
Words cant even describe thank you so much
Great teacher
I wish i can meet you one day
Big respect from iraq
Thank you so much! Perhaps one day we'll meet. Greetings from South Africa
Thanks a lot sir Michael 🎉
Very nicely explained as usual 👏
U r awesome 👌🏻
thank u i was seaching for all these information for days u just made it easyer❤❤
That's great news! So glad it was helpful!
Thank you! The video is very helpful, I just don't understand why you say that in order to convert the time domain into frequency domain we use an inverse FT instead of a FT. Thank you again! If anyone else knows the answer, that would also be helpful :)
Thankyou Micheal appreciate your efforts
How do you tell by looking at a mri image which is the frequency direction vs phase direction
Thanks ! please have you videos on The gradient matrix and concomitants fields?
thank you sir very much, these videos are helping me A LOT !!! actually I just might just study from these videos for the ARRT BOARD !!!
You are most welcome! Good luck for you exam 🤗
I have a question, could you please consider answering. At 6.39, you said the spins are spinning in-phase with one another, since the only magnetic field they experience is the main magnetic field. In the very first video you said and also showed with an animation that, with the influence of the main magnetic field, most of the spins will align parallel to the main magnetic field. They will precess at the same frequency (Larmor frequency), but not in-phase with one another.
I have the same doubt. if the spins thend to dephase the animation in sincrony confuses me so why the gradientes if each voxel have tissue with diferent decaiyng.
excellent video. solved a lot of my doubts, got few new doubts.😅
Yash bro brt se ho kiya
No. Md
Thank you doctor 😊
Thankyou micheal🎉🎉
Another question... it seems you overlay the frequency encoding chart with k-space when explaining how to get the final picture, is this accurate conceptually? Wouldn't this mean that a frequency encoding chart has a Y-axis? I thought that any frequency encoding basically generates rows all having the same frequency essentially just generating only more x-axis rows?
Thank you Dr. A question: It is written in Sarah that in repeat cycles of phase encoding gradient, each point along the segment will have DIFFRENT FREQUENCY. Those that are at furthest ends will have higher frequency phase encoding curve. Those frequencies will differ when we apply frequency encoding gradient? You mentioned that frequencies will be same. These will be same before applying frequency encoding gradient?
Great observation. There are two separate frequencies being talked about here. The frequency of precession along the x axis will remain the same for each acquisition. The frequencies mentioned in the phase direction refers to the 2D frequency waveform vectors formed by both the imaginary and real components of the spins in the entire slice - these are dependent on the combination and phase and frequency gradients. In this course I have largely omitted 2D waveforms in order to avoid confusion. In the k-space talk when I mention the rate of change of frequencies increasing at the periphery of k space, I am referring to these 2D phase frequencies. Hope this hasn't caused you more confusion 😅
@@radiologytutorials Thank you
Excellent job 😍😍
Thanks Sohail. I hope it made sense. Difficult one to explain without getting too detailed 😆
@@radiologytutorials yes sir too much detailed confuse us fewer concepts enough to made my day😃😃
these tutorialos are a lifesaver. Micheal can you do flouro aswell. i cant understand that shit for the life of me 😫
Great Job
Brilliant video thank you! One thing I didn't understand about the two diagrams around minute 30 (k space vs. after first Fourier transform): why does the diagram on the right not have rows with less and less intense signal at the top and the bottom? Those were acquired with a stronger gradient so I thought they'd have weaker signal?
At the 7th minute when you were explaining that the magnetization vectors within a slice were in phase with one another, I got confused. I thought all the vectors would be out of phase until they reach TE after the 180 degrees RF pulse, given that different tissues may exist within a selected slice?
Congrats great course Im almost there to get it. but please answer, why the gradientes are needed if the spins thend to dephase by themself by the t2 decaiyn .the video off the spins in sincrony confuses me. wont the dephasin itself would create a signal in y and x because each voxel have tissue with differente decaiyn rate?????
Hi there, I am trying to put together the bigger picture: the phase encoding gradient is applied to the entire slice, and spins in the phase direction change their rate of precession depending on their location embedding a phase shift, when we apply the 180 rephasing pulse at a certain point of relaxation, does it only rephase the spins that still have some transverse mag? How does the addition of the gradient change the echo if the spins all align anyway with rephasing?
My understanding so far is that with a certain amount of relaxation, we get different amounts of signal depending on tissue, I that we have an echo that is a bunch of different signal strengths based on T1 and T2 differences, but how does the change of frequency along the phase encoding gradient fit into this?
hello sir, do you turn on the FEG again after the PEG is used? why do you have to turn the PEG off?
Why do we have to take so many phase signals?
Why can't we just do FED in x-axis in one cycle?
and FED in y-axis in second cycle and this will give us y axis values.
Does phase encoding only help with getting y-axis signals?
Thank you for uploading these helpful tutorials! I am confused about why FEG changes the frequency and PEG changes the phase of the spins, considering they both induce gradient fields?
Great question. PEG also changes the frequency for the time the gradient is on. However, once the gradient is turned off the spins will continue to precess at larmor frequency but will be out of phase. The FEG changes frequency (and is applied whilst we are reading the signal - therefore that frequency change is registered in the signal). Hope that makes some sense 🤞🏼 let me know if you need more explanation 🙂
I see! Thank you so much! @@radiologytutorials
You explained things so clearly, I truly appreciate the help! @@radiologytutorials
Absolute pleasure! Glad it’s helpful 🙂
As I see it. The PEG gradient last for a short period of time, not enough as to cause a change at frequency that will translate to the signal.
can anybody please explain the advantages and disadvantages or a maxwell coil vs a golay coil?
thank u a lot! i kove your videos
As always it’s only a pleasure ☺️
I get a little confused when you say something like "256 different pixels." To me, a "pixel" is a final representation of an image voxel in the matrix. Like a camera picture, it's the smallest form of the grey scale which makes the image. Do you mean it more like "256 data points" like a numerical value of signal intensity? As an MRI tech, when I change the frequency matrix, I'm thinking I'm changing the number of frequencies I want to sample. For example, if the default frequency encoding number is 256, I think that means I'm telling the scanner to use a gradient and signal math to give me 256 different frequencies across the slice. The data gathered in that gives the scanner 256 differences to measure across the slice. And if I want to half that, I can choose 128 and now I get 128 different frequencies across the slice. It's like I'm literally choosing a data matrix across the slice. Am I thinking of this correctly? Or am I missing something about how K space is filled from the frequency encoding? TIA!
To clarify, I think of a 256 frequency matrix as dividing the slice along the frequency direction into 256 "columns" of different frequencies. Half of that would be 128 "columns" across the same space, meaning less frequencies are used across the slice to encode the exact same space the 256 was. This would mean we are creating 128 detectable differences in the slice that we can then use to associate with the known applied gradient to give us spatial location along a single axis. Dividing this info into pixels doesn't quite make sense to me, and I'm hoping you can help me understand. I'm also a bit confused as to how k-space is filled and cross-referenced with the frequency data. I thought frequency data was included in k-space. I haven't reached your video on that yet, so the answer to my question might very well already be out there!
Sorry for the delay in replying. Your thinking is correct.
Mostly these terms can be used interchangeably but there are subtle differences. Can't remember exactly how I phrased it in this talk (I may have caused some confusion). I try to use matrix size and pixels along the x axis as synonyms because both refer to the spatial resolution in the field of view. The matrix size determines the number of voxels and ultimately pixels we will display within the FOV.
The data acquisition sample number (ie the number of digital data points we sample from the analogue signal) determines the number of frequencies we can separate using a Fourier transform. Some will call this the frequency matrix. You're right, the number of times we sample a signal will determine the number of different frequencies we can delineate and ultimately place on an x axis location. However, there are sequences where we oversample the frequency encoding direction and disregard data that falls outside of the field of view (like when we oversample to avoid aliasing). In this case the frequency matrix and the image matrix do not correspond.
I may not have answered your question. If not, let's chat some more!
@@radiologytutorials I appreciate your reply , and thanks for telling me I'm kind of on the right track. I think what's hard about the pixel reference for me is that to encode 2 different locations within the same frequency, the phase encoding gradient has to come in to play to differentiate them. I think in the following video you clear all of this up pretty well.
The hardest thing for my students to get over is this thought that we're still recording an image in MRI. I literally have started telling them "none of this is real!" lol. When they really grasp that there is no picture, that our images are graphical representations of electrical signals, I think it really helps them start understanding how chemical shift, aliasing, and other artifacts are just electrical signals the computer can't process correctly. Just today I was having to scan at the edge of the magnet bore, and the computer has a really tough time encoding signals at that extreme of a gradient. The fat saturation was abysmal, and there was signal cut off. The scanner ran frequency encoding pre-scans for like 2 minutes trying to figure it out, and even then there were issues. It's a challenge, but it sure isn't boring!
How is it possible to a phase encoding gradient and not have that also create a frequency encoding gradient?
Whilst the phase encoding gradient is on it is essentially creating a difference in frequencies along the slice. However, when the gradient is switched off all the spins now precess at the same frequency because the only magnetic field is the main magnetic field. The spins now have different phase but the same frequency. This differs from the frequency encoding gradient because we apply that gradient whilst reading the signal - therefore these frequency differences will be in our data. Hope that helps 🙂
took me a few times reading it over but that now makes sense, ty@@radiologytutorials
basically its the on and off of the gradients thats key. Putting it on does change the frequencies along the Y-axis AND keeps the phases and then the turning off stores the phase but the frequencies all realign as they are experiencing a uniform magnetic field. Did I get this right?@@radiologytutorials
I think I need to buy knitting needles to understand this one :)
😂😂
this was tough T_T
k space doesnt have time as an axis
@@johnnysilverhand1733 k space is a frequency and phase domain. the output is decomposed into signal from each point in cartesian space
@@johnnysilverhand1733 thats only due to the time for relaxation of hydrogen atoms and local magnetic field effects avoided by not scanning evetything at once. the time component is not fundamental to the signal. without those factors, if we made the phase encoding steps 500ms, 5000ms, or 1 hour apart it will look the same as long as nothing in that region changed significantly