I’m a nuc med tech training in MR at the moment. I just want to say thank you. These lectures are brilliant. Coming from a nuc med background I felt really lost on the physics for MR. Now I finally understand how an image is formed.
Definitely the best explanation on TH-cam. Thank you for taking the time to explain this. It is indeed an inherently confusing concept and I'm happy that I was finally able to make the connection between phase encoding steps and populating the rows of k-space.
It is honestly revigorating to have such a nice and invested professor. Thank you for making MRI even more beautiful to me, and a lot easier to understand!
The very best MRI video (I had watched quite few of them) on TH-cam for a layperson to get some idea about how MRI was able to transform some electrical wave into image.
Best MRI physics EVER. You should expand the topics and build on those videos as is the best attempt to explain a rather complex topic that stops thousands of doctors to understand better that modality.
Thanks sir,Best video of mri physics...one question, is X,Y,Z gradient activated for localization and acquisition of echo signal in all sequence like SE, GRADIENT ECHO,IR & EPI or only during gradient sequence?
I'm glad you found this helpful but it is only meant for introductory role, certainly nowhere near as comprehensive as Dr. Lipton's series, which are fantastic. I'm a huge fan of his book and learned a great deal from him as a clinical radiologist. I'll try to answer as best as I can. Anyway, my e-mail is isaacwu@hotmail.com. Cheers.
I don't quite understand the graph at roughly 28:31... how does the frequency gradient change the amplitudes of the signals or in a sense the echo curve? I thought employing a gradient was to be able to localize a signal to a certain place along a dimension but wouldn't impact the signal? Like based on this, wouldn't their be signal loss within each slice due to the varying gradients which would impact contrast?
I am confused as to why we exactly use the Fourier transformation considering the "complex" aggregate signal is STILL transformed into an aggregate signal? Is the reason to create a signal with only positive amplitudes somehow?
This is an excellent (though difficult/confusing at times) tutorial on this subject. One part I wish you would have clarified is the type of gradients you were referring to. You talk about gradients but fail to mention whether it's frequency or phase encoding gradient (FE/PE), which can be a bit frustrating and confusing. I had to guess at times, thinking you were talking about the FE gradient after the slice selection. It'd be nice if was highlighted clearly often as being repetitive is a good thing, especially dealing with this complex topic. I guessed PE is only used if you wanted to move within the k-space matrix. Also, in the signal timing diagram, if you have included where the slice selection gradient occurs, it would have made the graphic clearer, I believe. It only included the frequency and phase encoding gradients timings, but maybe you had a reason why you drew them that way. So, please help me out here to see if I have things right... 1. You select a slice using Gss gradient and let's say you pick the 63MHz slice where all 16x16 matrix voxels within that slice will spin at 63Mhz, right? 2. Then set the phase Gpe gradient to point to the first row and now apply a frequency Gfe gradient, where the spins within that row now vary from 62.9 MHz - 63.1 MHz, for example,(left to right) due to the gradient. Then is the FE readout signal is a single time-domain signal representing these spins from 62.9MHz to 63.1MHz, right? Then you perform a 16-sample FFT of this signal fills the first k space row, right? 3. Then step down the phase Gpe gradient to point to the 2nd matrix row and repeat step 2 all over again until the whole 16x16 matrix is filled, right? 4. Once the full 16x16 k-space is completed, then you perform a 2D inverse FFT to get the real image of the original slice, right? 5. Repeat select another slice using a different Gss gradient and repeat steps 1-4. 6. So, at what steps (1-4 above) do you compute T1 and T2 for each individual voxel within the selected slice? This puzzles me ... I believe your presentation is excellent and could easily become THE reference MRI material on image localization for non-physicists like me (Biomedical Engineering PhD student) if you had included the clarifications I mentioned above. A lot of esoteric MRI gradient names are floating about in our minds, so highlighting which particular one you're referring to would help a great deal and leave out all guesswork. Thanks for the great time and effort you put into this presentation as it has helped me understand the localization process much better. The few questions I had for you would bring that home for me 100%. Thank you, again!
Hi. When I was a radiology resident, I found MRI physics an extreme tough and dry subject to understand. It’s an even tougher subject to teach now that I’m an attending. Thank you your suggestions; some really good ideas to incorporate into my future lectures for my residents. By the way, just a terminology thing, but the proton does not spin at 63 Mhz but rather precess at 63 Mhz. Analogy is like the earth is spinning (spin) on its axis but also revolving (precess) around the sun. The rate of precession can be calculate based on Larmor’s equation, depending on how strong the main magnetic field is. But I get what you mean. Anyway, you got the steps 1 to 5 pretty much right. I'm probably not understanding your question 6 correctly, but T1 and T2 values for hydrogen protons in a specimen (fat, water, soft tissue, etc.) are not computed during these steps. If you’re referring how to construct T1 weighted vs. T2 weighted imaging, then it has to do with manipulation of the parameters TR (repetition time) and TE (echo time). The TR affects T1 weighting more and TE for T2 weighting. If you make TR short, you’ll get T1 weighted image. If you make TE long, you'll get a T2 weighted image, etc. I hope that answer some of your questions. Thanks again.
@@neuroradish Thank you for the kind and fast reply to my earlier questions and yes, I should be more precise in my terminology regarding proton spin or precession to be exact. Now that you described my steps 1-5 as basically accurate(hope so?)... a. When does one compute the T1 and T2 values of each voxel within the image? Is it during the above steps 1-5 process or some other process at another time? b. Is a T1-weighted or T2-weighted image basically a map of the T1/T2 values contained within each voxel of the image, scaled to be between 0 -255 for an 8-bit image? I'm thinking that the T1/T2 values for fat, water, and soft-tissue are determined by simply pointing to the fat tissue, for example, and just reading off its T1/T2 values as an average over the entire selected region. I am thinking of a CT image, where each pixel is the x-ray attenuation coefficient for that part of the body, scaled from 0 to 255. Thanks again!
@@davidmusoke Of course; glad to help. Yes, you have the steps 1-5 correct. For question a) is that we don’t really compute T1 and T2 values. T1 or T2 values are already known under a certain condition / magnetic field. For example, at 1.5T magnetic field, T1 value for fat is approximately 250 (milliseconds), grey matter 900, CSF 4000, etc. Most MRI technologists (and radiologists) don’t really these values memorized. We just push some buttons and the computer usually sets the parameters for us. We just need to specify what type of images we want. Sometimes we do some tweaking if the images don’t turn out looking quite right. We do take advantage of these known values to help us differentiate between different tissues or pathology vs. normal tissue, based on their different T1 or T2 relaxation time. The most difficult part for someone like me with limited physics background to grasp is that MRI is constructed so differently than CT. To me, CT is much easier to visualized in term of its signal localization and how the final image should look like based on tissue density (and how much x-ray is attenuated by that tissue). Not unlike an analog photograph or plain x-ray. But for MRI, T1 weighted imaging for example, it utilizes the different T1 relaxation time between tissues to differentiate them. So on T1 weighted imaging, fat with a very short T1 (~250 ms) therefore faster T1 relaxation, would be hyperintense. In comparison, water/CSF (~4000ms) with its long T1 relaxation time, would have low signal intensity. The exact inner working of what value is assigning to certain gray scale is not something I can explain concisely, I’m afraid. But I have a feeling that I’m not answering your question completely. I’d like to suggest a great resource for MRI physics is (www.mriquestions.com), if you don’t already known. Some concepts are much more than I need for my clinical work, but I found this site extremely helpful for my learning and I think you’ll find it very helpful, too. Hopefully that helps.
@@neuroradish Yes, it does, and thanks for the resource you mentioned (mriquestions.com). It seems a rather in-depth list of answered questions for MRI technologists or radiologists. Thank you again for your time as you've been a great help!
In minute 26 you explained you need to apply multiple phase encoding steps why not a single one it’s enough to cause the whole rows from bottom to top to be phase shifted i hope u get my point what i mean if u apply single application its going to induce phase shift corresponding or depending on its location the one in top will have postive higher frequency more phase shifted than than the one bellow it because the field is getting weaker as u go down so the spins will experience less phase shift than above
Ahaha. I didn’t even put it together. Thanks for that. All I could think of was that movie where the person turned off the MRI so he could escape the magnetic field in time. 🤦♂️
Very good explanations generally! But i have a question: in the second step (18:15) to acquire the k space image when i turn on the frequency gradients,doesnt the phase change ??
Hi. Thank you for watching. You're right, the phase also change whenever we turn on a magnetic gradient. In this case, the gradient needs to stay on in order to cause a frequency shift, so we can use different frequencies for localization (between columns in our example). Since we are trying to detecting frequency difference, phase shift would not alter our ability to do that. But because there is different phases, we've also caused more de-phasing of the spins, and therefore increased signal loss. To compensate for that signal loss, we can do a frequency gradient in the opposite polarity first, then do our intended frequency read out gradient. In other words, we purposely de-phase the spins first, then during the read-out (frequency encoding), we essentially re-phase these spins to regain the signal loss. I found this graph at this website ( mriquestions.com/what-is-fsetse.html ). I find this website to be extremely useful for MRI physics. Notice in the diagram, the Freq gradient has 2 pulses, one before 180 RF pulse, and one after for sampling echo (Signal). By the way, the reason the first pulse looks like it's in the same polarity as the 2nd pulse (instead of drawing it upside down) is because of the 180 RF pulse. I hope this helps.
@@neuroradish Thank you for your time to answer! So by this answer ,in a nutshell you mean that by shifting the frequencies we have a shift in phase too .This change doesnt not stop us from applying the frequency gradient but it affects the signal amplitude.With a technique mentioned we regain our weak signal. Correct me if i am wrong!!Thank you again and stay safe! My greetings from Crete,Greece!
When you increase magnetic gradient, you have "sped" up the phase incoherence between protons in the transverse plane. The difference between the precession frequencies of the protons would increase, that would lead to faster decay of transverse magnetization, therefore faster signal loss.
For to do the wavelet transform or Function Wave i am not requiere the complex number or Fourier series. Just I need the new methodoly I discovered.I left this video to compare: th-cam.com/video/3Ebvypj577E/w-d-xo.html
!["ซี - เอมี่ " ย้อนความรัก 19 ปี ที่ต้องทนกันและกัน... | แฉ 22 ต.ค. 67 [2/3] | GMM25](http://i.ytimg.com/vi/ykv9ILzatHM/mqdefault.jpg)
I’m a nuc med tech training in MR at the moment. I just want to say thank you. These lectures are brilliant. Coming from a nuc med background I felt really lost on the physics for MR. Now I finally understand how an image is formed.
Definitely the best explanation on TH-cam. Thank you for taking the time to explain this. It is indeed an inherently confusing concept and I'm happy that I was finally able to make the connection between phase encoding steps and populating the rows of k-space.
It is honestly revigorating to have such a nice and invested professor. Thank you for making MRI even more beautiful to me, and a lot easier to understand!
No one was able to help me with this part with this clarity. Thank you so much.
the best explanation for MRI physics for radiologists that Ive seen, thanks so much!
yeah but do respect your Technologist that has much more understanding about the physics than you?
The very best MRI video (I had watched quite few of them) on TH-cam for a layperson to get some idea about how MRI was able to transform some electrical wave into image.
The best illustration I have ever seen
Best MRI physics EVER. You should expand the topics and build on those videos as is the best attempt to explain a rather complex topic that stops thousands of doctors to understand better that modality.
The illustrations are amazing!
One of the best MRI lectures I've ever seen here
He says stronger the gradient more the signal loss what does that mean?
the previous video of this serie was incredible and this one too... with the frequency encoding and phase encoding... amazingg
One of the best explanation videos for MRI principles, thank you so much!
Absolutely amazing explanation! Thank you so much (neuroscience student)
Amazing.. Finally I can grab the concept... Thanks very much..
Really cool explanation, my favorite part is the cyborg cat doll :p
For beginners, this video speaks very clearly!!
So, a person who is in the MR scanner must be very excited!😆
Amazing. Now i can go confidently to attempt my MRI exam
Bestest lecture ever I've found🤩🤩👍👍👍
Thank you soo much sir
Amazing video, Sir. Thank you so much for the lecture. It helps me a whole lot.
Crisp and amazing! thank yo so much sir...
Really great work, thanks
Excellent illustration... Thank you sir....
Amazing it all comes together :)
You are Amazing
this is an incredible video. well done sir
Thanks sir,Best video of mri physics...one question, is X,Y,Z gradient activated for localization and acquisition of echo signal in all sequence like SE, GRADIENT ECHO,IR & EPI or only during gradient sequence?
Hi. That's correct, depending on which direction(s) / location (s), it is applied for all, not just for gradient echo. Thanks
Hi sir , Please make ASL,T1 & T2 perfusion, Functional MRI, TRICKS,MOTSA and Cardiac MRI physics video also.
@@neuroradish Thanks for your reply sir
brilliantly simplified. 👏
thanks very much for the clear explanation.
Love your explanations. Thank you!
Better than dr.michael lipton's lengthy lecture.superb,concise yet comprehensive.can you provide your email for doubt clearance?
I'm glad you found this helpful but it is only meant for introductory role, certainly nowhere near as comprehensive as Dr. Lipton's series, which are fantastic. I'm a huge fan of his book and learned a great deal from him as a clinical radiologist. I'll try to answer as best as I can. Anyway, my e-mail is isaacwu@hotmail.com. Cheers.
Keep up the good work 👌👌👌
Finally, I get it. thank you!
Thank you
I don't quite understand the graph at roughly 28:31... how does the frequency gradient change the amplitudes of the signals or in a sense the echo curve? I thought employing a gradient was to be able to localize a signal to a certain place along a dimension but wouldn't impact the signal? Like based on this, wouldn't their be signal loss within each slice due to the varying gradients which would impact contrast?
I am confused as to why we exactly use the Fourier transformation considering the "complex" aggregate signal is STILL transformed into an aggregate signal? Is the reason to create a signal with only positive amplitudes somehow?
Fantastic!!!!
The best expansion
BRILLIANT!!!
I love you sir,
Generally, we use more frequency encoding steps than phase. So how phase encoding > frequency encoding ?
This is an excellent (though difficult/confusing at times) tutorial on this subject. One part I wish you would have clarified is the type of gradients you were referring to. You talk about gradients but fail to mention whether it's frequency or phase encoding gradient (FE/PE), which can be a bit frustrating and confusing. I had to guess at times, thinking you were talking about the FE gradient after the slice selection. It'd be nice if was highlighted clearly often as being repetitive is a good thing, especially dealing with this complex topic. I guessed PE is only used if you wanted to move within the k-space matrix.
Also, in the signal timing diagram, if you have included where the slice selection gradient occurs, it would have made the graphic clearer, I believe. It only included the frequency and phase encoding gradients timings, but maybe you had a reason why you drew them that way.
So, please help me out here to see if I have things right...
1. You select a slice using Gss gradient and let's say you pick the 63MHz slice where all 16x16 matrix voxels within that slice will spin at 63Mhz, right?
2. Then set the phase Gpe gradient to point to the first row and now apply a frequency Gfe gradient, where the spins within that row now vary from 62.9 MHz - 63.1 MHz, for example,(left to right) due to the gradient. Then is the FE readout signal is a single time-domain signal representing these spins from 62.9MHz to 63.1MHz, right?
Then you perform a 16-sample FFT of this signal fills the first k space row, right?
3. Then step down the phase Gpe gradient to point to the 2nd matrix row and repeat step 2 all over again until the whole 16x16 matrix is filled, right?
4. Once the full 16x16 k-space is completed, then you perform a 2D inverse FFT to get the real image of the original slice, right?
5. Repeat select another slice using a different Gss gradient and repeat steps 1-4.
6. So, at what steps (1-4 above) do you compute T1 and T2 for each individual voxel within the selected slice? This puzzles me ...
I believe your presentation is excellent and could easily become THE reference MRI material on image localization for non-physicists like me (Biomedical Engineering PhD student) if you had included the clarifications I mentioned above. A lot of esoteric MRI gradient names are floating about in our minds, so highlighting which particular one you're referring to would help a great deal and leave out all guesswork.
Thanks for the great time and effort you put into this presentation as it has helped me understand the localization process much better. The few questions I had for you would bring that home for me 100%. Thank you, again!
Hi. When I was a radiology resident, I found MRI physics an extreme tough and dry subject to understand. It’s an even tougher subject to teach now that I’m an attending. Thank you your suggestions; some really good ideas to incorporate into my future lectures for my residents.
By the way, just a terminology thing, but the proton does not spin at 63 Mhz but rather precess at 63 Mhz. Analogy is like the earth is spinning (spin) on its axis but also revolving (precess) around the sun. The rate of precession can be calculate based on Larmor’s equation, depending on how strong the main magnetic field is. But I get what you mean.
Anyway, you got the steps 1 to 5 pretty much right.
I'm probably not understanding your question 6 correctly, but T1 and T2 values for hydrogen protons in a specimen (fat, water, soft tissue, etc.) are not computed during these steps. If you’re referring how to construct T1 weighted vs. T2 weighted imaging, then it has to do with manipulation of the parameters TR (repetition time) and TE (echo time). The TR affects T1 weighting more and TE for T2 weighting. If you make TR short, you’ll get T1 weighted image. If you make TE long, you'll get a T2 weighted image, etc.
I hope that answer some of your questions. Thanks again.
@@neuroradish Thank you for the kind and fast reply to my earlier questions and yes, I should be more precise in my terminology regarding proton spin or precession to be exact. Now that you described my steps 1-5 as basically accurate(hope so?)...
a. When does one compute the T1 and T2 values of each voxel within the image? Is it during the above steps 1-5 process or some other process at another time?
b. Is a T1-weighted or T2-weighted image basically a map of the T1/T2 values contained within each voxel of the image, scaled to be between 0 -255 for an 8-bit image? I'm thinking that the T1/T2 values for fat, water, and soft-tissue are determined by simply pointing to the fat tissue, for example, and just reading off its T1/T2 values as an average over the entire selected region.
I am thinking of a CT image, where each pixel is the x-ray attenuation coefficient for that part of the body, scaled from 0 to 255.
Thanks again!
@@davidmusoke
Of course; glad to help. Yes, you have the steps 1-5 correct.
For question a) is that we don’t really compute T1 and T2 values. T1 or T2 values are already known under a certain condition / magnetic field. For example, at 1.5T magnetic field, T1 value for fat is approximately 250 (milliseconds), grey matter 900, CSF 4000, etc. Most MRI technologists (and radiologists) don’t really these values memorized. We just push some buttons and the computer usually sets the parameters for us. We just need to specify what type of images we want. Sometimes we do some tweaking if the images don’t turn out looking quite right. We do take advantage of these known values to help us differentiate between different tissues or pathology vs. normal tissue, based on their different T1 or T2 relaxation time.
The most difficult part for someone like me with limited physics background to grasp is that MRI is constructed so differently than CT. To me, CT is much easier to visualized in term of its signal localization and how the final image should look like based on tissue density (and how much x-ray is attenuated by that tissue). Not unlike an analog photograph or plain x-ray.
But for MRI, T1 weighted imaging for example, it utilizes the different T1 relaxation time between tissues to differentiate them. So on T1 weighted imaging, fat with a very short T1 (~250 ms) therefore faster T1 relaxation, would be hyperintense. In comparison, water/CSF (~4000ms) with its long T1 relaxation time, would have low signal intensity. The exact inner working of what value is assigning to certain gray scale is not something I can explain concisely, I’m afraid.
But I have a feeling that I’m not answering your question completely. I’d like to suggest a great resource for MRI physics is (www.mriquestions.com), if you don’t already known. Some concepts are much more than I need for my clinical work, but I found this site extremely helpful for my learning and I think you’ll find it very helpful, too.
Hopefully that helps.
@@neuroradish Yes, it does, and thanks for the resource you mentioned (mriquestions.com). It seems a rather in-depth list of answered questions for MRI technologists or radiologists. Thank you again for your time as you've been a great help!
In minute 26 you explained you need to apply multiple phase encoding steps why not a single one it’s enough to cause the whole rows from bottom to top to be phase shifted i hope u get my point what i mean if u apply single application its going to induce phase shift corresponding or depending on its location the one in top will have postive higher frequency more phase shifted than than the one bellow it because the field is getting weaker as u go down so the spins will experience less phase shift than above
CLEARLY Arnold is here to remind us that in Terminator 1 (T1), he's bad (fluid is dark) and in Terminator 2 (T2), he's good (fluid is bright)!
Ahaha. I didn’t even put it together. Thanks for that. All I could think of was that movie where the person turned off the MRI so he could escape the magnetic field in time. 🤦♂️
nice lecture thanks
This is a great lecture!
Thanks a lot
Very good explanations generally!
But i have a question: in the second step (18:15) to acquire the k space image when i turn on the frequency gradients,doesnt the phase change ??
Hi. Thank you for watching. You're right, the phase also change whenever we turn on a magnetic gradient. In this case, the gradient needs to stay on in order to cause a frequency shift, so we can use different frequencies for localization (between columns in our example). Since we are trying to detecting frequency difference, phase shift would not alter our ability to do that. But because there is different phases, we've also caused more de-phasing of the spins, and therefore increased signal loss. To compensate for that signal loss, we can do a frequency gradient in the opposite polarity first, then do our intended frequency read out gradient. In other words, we purposely de-phase the spins first, then during the read-out (frequency encoding), we essentially re-phase these spins to regain the signal loss. I found this graph at this website ( mriquestions.com/what-is-fsetse.html ). I find this website to be extremely useful for MRI physics. Notice in the diagram, the Freq gradient has 2 pulses, one before 180 RF pulse, and one after for sampling echo (Signal). By the way, the reason the first pulse looks like it's in the same polarity as the 2nd pulse (instead of drawing it upside down) is because of the 180 RF pulse. I hope this helps.
@@neuroradish Thank you for your time to answer!
So by this answer ,in a nutshell you mean that by shifting the frequencies we have a shift in phase too .This change doesnt not stop us from applying the frequency gradient but it affects the signal amplitude.With a technique mentioned we regain our weak signal.
Correct me if i am wrong!!Thank you again and stay safe!
My greetings from Crete,Greece!
@@dimitheodoro Yes, that's correct. You stay safe as well. Thanks.
I just read frequency encoding is applied continuously whereas phase encoding is discontinuous,so how can you get 16 frequency steps?
Why do we have to suppress signals in the three rows mentioned?
What the peak of free induction decay represents.? Why its higher amplitude ? Because of what
What is signal loss ?how does it increase when you increase the gradient?
When you increase magnetic gradient, you have "sped" up the phase incoherence between protons in the transverse plane. The difference between the precession frequencies of the protons would increase, that would lead to faster decay of transverse magnetization, therefore faster signal loss.
had-fhadh video :)
For to do the wavelet transform or Function Wave i am not requiere the complex number or Fourier series. Just I need the new methodoly I discovered.I left this video to compare:
th-cam.com/video/3Ebvypj577E/w-d-xo.html