Transcript for:
Exploring MRI: Boltzmann Magnetism and Spin Echo

This is the second video in a small series  discussing how MRI works - delivered in   a very timely fashion. If you haven't seen Part 1,  you really should watch that first or you may get   lost here - link in the description. In this video  I'm hoping to go a little deeper into Boltzmann   Magnetism and our magnetic moment ensemble, the  NMR experiment, and the effects of inhomogeneous   magnetic fields. This will all lead us to the  primary topic of the video: the spin echo. On our last episode we discussed the basics of  the NMR phenomenon including precession, T1 and   T2 relaxation, and the signal equation following a  90 degree pulse. Then we asked the question: what   is Boltzmann Magnetization., M0? Well we touched  on it last time, but never actually defined it.   Instead we examined a related quantity: the  polarization - which describes the fraction of   spins which contribute to NMR signal. So lets see  how polarization and magnetization are related. Continuing with where we left off in the last  video, Boltzmann magnetism or magnetization   is the equilibrium nuclear magnetism of a spin  sample in a magnetic field. Every proton has the   same magnetic moment, mu which is equal to gamma  (it's gyromagnetic ratio) times hbar (the reduced   planks constant) times its spin quantum number  which in the case of protons is 1/2. As we know,   our sample is composed of a great many protons  each with the same magnetic moment. Boltzmann   magnetism is simply the vector sum of all of these  magnetic moments when they are in equilibrium.   Of course, in the absence of a magnetic field  these spins lack a defined orientation, but   in the presence of the magnetic field, these  spins orient themselves in 1 of 2 possible   states - spin up or spin down. The spin up state  (where the spin is aligned with the field) is the   low energy state and is thus preferred, but as we  saw in the last video, the energy difference is so   small compared to the temperature of the sample  that the spins are nearly equally distributed   between these states - only a few spins out of  every million are in excess in the spin up state. In the last video, this led us to a discussion of  polarization: the Fraction of spins which are in   excess in the aligned, low energy state - which  we saw simplified to this expression for samples   in thermal equilibrium - as a reminder this  polarization is the spins' gyromagnetic ratio   gamma, times the reduced planks constant hbar,  times the external magnetic field strength B0,   divided by 2 times boltzmann's constant  k times the sample temperature T   in kelvin. We were able to work this out directly  using concepts from Boltzmann statistics (again,   link in the description for a refresher)...but  then we just kinda stopped without actually   defining Boltzmann Magnetism. But here's  a bit more detail to help out. If we want   to actually calculate Equilibrium  magnetization of our spin sample,   like we said, that should just be the vector sum  of our magnetic moments. Since we know that the   spins exist in only 2 states - up or down - we  know every pair of up and down spins will result   in a net cancellation of magnetization. Thus  only the unpaired spins contribute to sample's   nuclear magnetism. And now we can intuitively  write an expression for Boltzmann magnetism:   it's simply the magnetic moment of a single spin  times the number of spins we have in our sample,   then times the polarization (the fraction  of those spins which are unpaired). Plugging in these values we find  the total magnetization of the   sample is given by this expression for protons  (ie hydrogen nuclei) in a magnetic field B0.   We often compile all of these terms  into a single variable, chi naught,   which is the static nuclear susceptibility  of a sample - it essentially tells you how   big of a magnetic moment you'll get in a given  magnetic field strength - and it is inversely   proportional to temperature. This is called 'curie  law' (after Pierre Curie, husband of Marie Curie).   As we've said a number of times Boltzmann  magnetism is the nuclear magnetism of our   sample in thermal equilibrium, so I find  myself using those terms interchangably:   Boltzmann magnetization/equilibrium magnetization.  If you want to test your understanding a bit   further, work out the units of this expression -  does it match the SI unit of a magnetic moments? With this foundation, I'm going to abandon the  Energy state picture of nuclear magnetism and   replace it with a single vector representing  our Boltzmann Equilibrium magnetization.   Since we always align the z axis of our frame  of reference with the magnetic field B0,   our Boltmann magnetization vector  is created along the same axis.   Any NMR experiment essentially involves knocking  our sample out of equilibrium whereupon it   delivers detectable signal as it returns  to Boltzmann - which we detect and analyze.   Here I'm illustrating an NMR experiment where  we might knock the nuclear magnetism our of   alignment with the B0 field by flip angle  theta. The key to understanding NMR dynamics   is that our magnetization vector will return to  Boltzmann following the relaxation principles we   discussed in the last video: it's transverse  magnetization will decay to 0 at the T2 rate   and its longitudinal magnetization will  return to Boltzmann at the T1 rate.   Because of the precession phenomenon, the amount  of our magnetization in the transverse plane   dictates the amount of signal we detect, which  we saw last time increases as the precession   frequency omega increases. And thus we report  an expression describing NMR signal of a sample:   the signal is proportional to the number of spins  in our sample times the magnetic moment of an   individual spin, times the fraction of  those spins contributing to the NMR signal,   times omega, which is gamma B0. Notice the  effect our magnetic field strength B0 has our   detected signal: Let's say we double our external  field strength B0, what happens to our signal?   It quadrupels! Remember for samples in thermal  equilibrium, there's another B0 buried in the   polarization expression. Increasing our magnetic  field increases our precession frequency AND our   Boltzmann magnetism - thus signal is proportional  to B0 squared. Neato Huh! Note however, that if   our sample is not in thermal equilibrium, such as  in hyperpolarized gas mri, then the expression we   derived for polarization no longer applies and our  signal will scale linearly with field strength. Before moving on, all of these magnetization  dynamics are typically summed up into a   tidy expression called the Bloch equation  named after Felix Bloch, pioneer of NMR.   I'm not going to go into details about it  here because there's other stuff to do,   but it would be remiss not to at least mention  such a fundamental equation here. Perhaps, in a   future video we'll build and solve the  Bloch equation. In its simplest form,   the Bloch equation says that in the presence  of a magnetic field B, a magnetization vector   M will precess about B in sort of conical  trajectory whose cone angle is essentially   twice the angle between B and M. In the mean  time, try working out the math for yourself! Ok, this may be a good time to clarify something.   On the last video I drew this ball and arrow  representation of a proton, which I think   helps to conceptualize the spin magnetic moment,  precession, flip angle, and relaxation phenomena.   Just to clarify: individual protons can only exist  in one of 2 Zeeman states within a magnetic field:   spin-up or spin-down. The dynamics we're examining  pertain to ensembles of many, many protons.  For example, the longitudinal magnetization  of our sample (ie the magnetization pointing   in the direction of our field), is actually  representative of the number of protons in   each state: higher if more spins are in the  spin-up state, negative if more spins are in   the spin-down state, and zero if there are  equal numbers of spins in each state. Thus,   longitudinal magnetization returning to Boltzmann  equilibrium, is simply the return of spins to   their equilibrium state populations which we saw  was contained within the polarization equation.   The signal detected during precession is the  energy released during state transitions. For   those familiar with electronic enegy transitions  which occur in fluorescent lamps and lasers,   this nuclear energy transition follows  nearly identical quantum mechanics.   The difference here is that the energy  separation of the two states is proportional   to the external magnetic field strength  B0, as opposed to just atomic structure.   Someday we'll do a proper video of NMR quantum  mechanics, but for now we'll stick with the   tried and true classical description. From  now on I will use a single arrow vector to   describe the nuclear magnetization vector  of our spins within a particular space. Ok, so the simplest NMR experiment we can devise  is to place a sample of spins in a magnetic field,   B0, and deliver a separate magnetic  field B1 perpendicular to B0 about   which our magnetization will precess. In a  reference frame which rotates at the larmor   frequency of our magnetization M in B0, the  B1 pulse is stationary. In the laboratory   frame of reference the B1 pulse rotates at the  larmor frequency, as does the magnetization M. When our magnetization has been rotated say 90  degrees it is entirely perpendicular to the static   B0 field maximizing our detectable signal during  its precession. We then detect the resultant free   induction decay as the magnetization works its  way back to Boltzmann equilibrium. This is often   called a 'ping' experiment or pulse/acquire. Note  that my vector cartoon is showing the rotating   frame picture - right, so it's animated in a  frame of reference that rotates at the larmor   frequency - but my Transmit and Receive plots here  show both the rotating frame signal (in white)   and the laboratory frame signals in yellow for  transmit and cyan for receive. We'll discuss   the mathematics behind transitioning between  rotating and laboratory reference frames in a   later lecture, but for the curious this is most  often achieved electronically by heterodyning. Oh, and notice that my rotating frame cartoon  shows B1 along the x hat axis which excites our   magnetization into the y hat axis. Turns out  you can precisely deliver B1 pulses along any   direction within the rotating frame's x-y plane  and thus place M along any direction you'd like.   In order to do this, you need to take note of  the phase of the larmor oscillator - the phase of   the oscillator within your excitation pulse will  dictate the excitation axis for your magnetization   and thus the phase of your detected signal as  well. If we look at our cartoon immediately at the   point in time immediately following the B1 pulse,  we can see that the choice of laboratory frame   phase in our pulse diagram dictates the precise  alignment of our B1 field within the rotating   frame and thus the detected signal. In the ping  experiment it may not seem like that matters too   much, but we'll see later that many NMR/MRI pulse  sequences are designed to leverage this fact,   especially in cases of multiple B1 excitations  so just keep that in mind as we move forward. So in order to perform this experiment IRL you  go to the NMR physics store and buy yourself an   NMR - a device which can deliver these  excitation pulses and display the acquired   signal on your oscilloscope. You prepare a  sample of plain water (plenty of protons - ie   hydrogen nuclei - to examine). You place  your sample in a strong magnetic field B0   and wrap the sample with wire which will be an  rf coil for transmission of excitation pulses   and subsequent reception of NMR signals. The NMR  is plugged into your oscilloscope and you window   the display to show 1 second of received signal  because you know from a simple google search   that the T2 of water is about 1 second, so you're  expecting the received signal to look something   like this - after 1 second the signal will have  decayed to about 1/e or 37% of its initial value. In exhilerating anticipation you execute the ping  sequence in your NMR AAND.....you get this....  You triple check your equipment, read all the  manuals, retry the experiment, but for some reason   the NMR signal you detect decays much  much more quickly than you expected.   Why does your signal decay so fast??? Perhaps your  water sample is contaminated with minerals which   reduce T2? Or perhaps the internet was wrong  when it said water has T2 of about 1 second?   Well actually it Turns out there's an  assumption we've made throughout our   tour of NMR which is now being challenged:  our B0 field isn't perfectly homogeneous.   We've assumed that B0 at every point in space is  exactly the B0 field strength in the z direction,   but of course no field is so  perfect, and even slight variations   in field strength and direction can have a  substantial effect on our NMR experiments. Consider two proton spins: One experiences a B0  field of 1 Tesla and the other experiences a B0   field of 1.000001 Tesla - exactly 1 microtesla  higher than his buddy. Given that the gyromagnetic   ratio of protons is 267.5 times 10^6 radians  per tesla per second (that's 42.6 MHz/T)   how long will it take before these spins are 180  degrees our of phase with each other and thus   will produce no NMR signal? Pause the  video to work it out if you'd like.   This difference in B0 leads to a difference  in omega following the Larmor equation.   The difference in B0 field is one millionth of  a tesla, and thus the difference in frequency   is 267.5 radians (or 42.6 revolutions) which  means the spin in the higher field will complete   42.6 more precession cycles than the first  spin every second. The spins will therefore be   pi radians or 180 degrees out of phase in  11.7 milliseconds! A difference in magnetic   field of one part per million reduces the  coherence of spins down to milliseconds! This is what's happening in our NMR  experiment - tiny imperfections in   our magnetic field homogeneity can have  a dramatic effect on our detected signal!   This is a problem, because now the  rate at which our signal decays does   not reflect properties of the sample - it  reflects the external conditions of our   experimental setup - namely how good (or  bad) of a magnetic field we can produce. So last we saw that signal decays at rate T2 due  to spin-spin interaction s, but when relaxation   due to field inhomogeneity is considered as well  our observed signal decays according to a new time   constant which we call T2*. That means our ability  to contrast tissue is substantially reduced.   So now what? Can we fix this? Oh, and you  might also be thinking 'wait a minute,   we saw on the last video that spins will  only experience excitation if the frequency   of our B1 excitation pulse is matched to the  spins' larmor frequency - well if the spins   are in different field strengths then they have  different Larmore frequencies. Could you still   excite both spins with a single-frequency B1  pulse? I guess their larmor frequencies are   close enough? what if the second spin was in a B0  field of 1.001 Tesla - a part per thousand off?'   If you're thinking along those lines, good!  We will absolutely examine all the intricate   details of excitation - in a later video.  For now we will assume all spins magnetic   moments experience excitation - even if their  larmor frequencies are not exactly equal. Let's examine explore this more closely in our  sample of spins here - I'll draw a single plane   of spins to make it easier to see, but note  that of course spins exist in the full 3D   space of our sample. I'll go ahead and add  a coordinate system to orient ourselves. We begin with a spin sample which is initially  in equilibrium with a magnetic field pointing   in the positive z direction - that is, they are  all at Boltzmann magnetizm. We'll work in the   rotating frame here - remember this means we  spin ourselves around at the larmor frequency,   so from this perspective the spins do  not precess about the B0 field - which   essentially means I won't animate the  spins precession about the z axis here. Ok,   let's deliver our B1 excitation magnetic  field in on the negative y axis. Our spins   then precess about the B1 field, and when  they are perpendicular to the B0 field,   i.e at a 90 degree flip angle, we turn off the B1  field. Now the total magnetization of our spins is   in the positive x direction of the transverse  plane and thus delivering detectable signal,   so let's think about the dynamics of  their subsequent dephasing and T2 decay.   I'll orient the camera so the magnetic field  is coming out of the screen here, add I'll   also add in another vector here to represent  the magnetic moment vector sum here as well. As we said in the last video, the randomly  distributed spins are experiencing the   external B0 field of our magnet, but also the  magnetic fields of each of their neighbors. If we   try to describe this for a single spin, we would  say its precession frequency, omega sub k for the   k'th spin, is equal to the larmor frequency plus  some Delta omega which of course is just gamma   times some delta B. And remember even though I'm  not animating it here, the spins are also bouncing   around all over the place by Brownian diffusion  so the delta B here can fluctuate in time as well.   This Delta B is what causes our spins to dephase  and our transverse magnetization to decrease   over time - the T2 decay we know and love. Now  we're working in the rotating frame, right,   which means we're spinning our frame of reference  at omega 0 - thus we drop the omega 0 term here.   The spins only exhibit precession in this frame  if their delta omega is non-zero - clockwise if   it's greater than 0, counterclockwise if its  less than 0. So let's get our magnetization   plot ready and play the animation. Hopefully  this seems familiar from the last lecture:   the spins dephase and total magnetization  decreases following e to the minus t/T2.   Rotating frame's kinda nice isn't it? Conveys all  the information we need without making me dizzy. Now most textbooks will illustrate  this using a 'stacked' view   of the spins rather than a messy image  of all the spins all over the place.   Since the NMR coil is assumed to detect all spins  with equal sensitivity this illustration makes   perfect sense and is much cleaner to visualize,  especially for our upcoming discussion on echoes.   So lets put our total Magnetization  vector back in and play the T2 animation.   This 'fanning out' of spins represents T2  decay which is manifest in the reduction   of the Magnetization vector length over time. With this picture available to us, perhaps  we can develop a mathematical expression for   the total magnetization vector M. This  shouldn't be too bad, right, after all,   the magnetization M is just the vector sum of  the individual magnetic moments mu sub k. Hmm,   ok, so what is mu sub k - the vector  describing the kth magnetic moment?   Well it's probably easiest to describe  in terms of its frequency: omega sub k.   If we do that we can see that the x-component  of mu is cos( omega sub k times time)   and the y component would be minus sin (omega  sub k times time). So that makes sense and it   suggests that if omega is positive, mu will  precess clockwise in this orientation which is   indeed how spins with positive gyromagnetic ratio  - like protons - precess about magnetic fields.   Well, I don't know about you but whenever  I see cos and sin arranged like this in a   vector expression, I immediately think of Euler's  relation. With this, we can abandon the Cartesian   x hat and y hat expression in favor of real and  imaginary axes. Thus I can express my vector as   a complex number e to the minus i omega sub k t.  This is so much cleaner because complex numbers   by design contain two pieces of information  - real and imaginary - in a single expression   with the added bonus that rotations are  much easier to describe mathematically.   If you're not used to doing this I would encourage  you to give it a try next time you're dealing with   rotations or sinusiodal motion - it truly makes  life much easier. Now, with each moment being   mu times e to the minus i omega t, we express  our transverse magnetization in time as such. Note that in this model, the individual spin  magnetic moments all have the same magnitude.   This makes mu a constant here so we can pull  that out front. Now summing over omega sub k   isn't super helpful unless we know what all the  mu's frequencies actually are. Well what we can   do then is examine the distribution of spins  with a given frequency - that is, the number   of spins which have a given omega - kind of like a  histogram for omegas but since it's continuous we   call it a frequency distribution (and the title  omega frequency distribution is not redundant,   this is a frequency distribution of omegas).  Anyways, Maybe it looks something like this:   most spins are pretty darn close to an omega  of 0 (which, in the laboratory frame would be   the larmor frequency). Some spins are a little  faster, some are a little slower, but as we get   further and further from w=0 the number of spins  with those frequencies goes down substantially.   I wonder what kind of distribution this is?  Well it actually doesn't matter, the point   is that this distribution, f(omega), describes  how many spins have a given frequency: at there   are f(omega 1) spins with frequency omega 1,  omega 2 would have f(omega 2) spins etc, etc.   So we can just rewrite our sum over k omegas as a  sum over all omegas where e^i omega t describes a   particular spin frequency omega, and f(omega)  is how many spins we have at that frequency.   Now of course we're dealing with septillions of  magnetic moments over a continuous spectrum of   possible frequencies - so this sum is really  an integral - and we have the result that our   Magnetization as a function of time (and thus our  detected signal as a function of time) is just   mu times the integral over all frequencies w   of the frequency distribution times  e to the minus i omega t d omega. O. M. G. This is the definition of the Fourier   Transform of the frequency distribution, f(w)!  This means that the rate at which our signal   decays in time, reflects the distribution of  frequencies our spins possess - and vice versa.   The way I drew this frequency distribution  here was not an accident - it is called a   Lorentzian distribution, and we'll do a deep  dive into the actual math in a later lecture.   My dear friends, Fourier Theory is THE  mathematical foundation of MRI - truly, you cannot   say you understand how MRI works if you can't  also say you know how the Fourier Transform works.   For some of you, the beauty of seeing  a Fourier transform appear here makes   perfect sense and you probably saw it coming  from miles away, but perhaps for others just   the phrase 'Fourier Transform' incites feelings  of dread and panic. Fear not! The next video in   our journey through MRI will be dedicated entirely  to what the Fourier Transform is and how it works. Until then, let's just develop a conceptual  understanding here. A sample of slowly dephasing   spins demonstrates a corresponding slow decay of  Magnetization and a narrow frequency distribution   showing a relatively tight grouping of spins  near omega=0 compared with more rapid dephasing   and quicker decay. This forms the basis  of T2 contrast weighting we saw last time.   Now because we know that omega equals gamma B0, we  know our omega distribution is actually indicative   of a distribution of regional magnetic field  B since all protons have the same gamma.   This distribution of the magnetic field strengths  experienced by our spins dictates our FID decay. So returning to our picture of the many spins  in space - we know that each spin experiences   some delta B due to the tissue environment, but  also some delta B due to our imperfect field. To   help us visualize, I'm going to show the magnetic  field in space as a color map. Spins in the purple   region experience a slightly lower field, and  spins in the red region a slightly higher field   and a full range of delta B in between.  The larmor equation tells us that field   inhomogeneity will broaden our frequency  distribution, thus reducing our decay time. So instead of placing our spins on a field  color map, I'll just color each spin according   to it's experienced field strength here. Now  we can stack them and observe the accelerated   dephasing. We'll use this stacked picture again in  a minute, but let's walk through the math a bit. Immediately following excitation all our spins  are aligned in the x direction and we have a   macroscopic ensemble moment to detect. If our  spins were all at the exact same frequency,   our signal would never decay and our  distribution would be perfectly narrow.   Our signal in time would just be a constant.  But of course our spins do dephase due   to the intrinsic molecular dynamics of our  system or tissue and dephase - our signal   is thus modulated by e^-Rt. R is the decay  rate and is characterized by time constant   T2 which we've seen before. Our signal  thus decays and our distribution widens. Now we have another source of decay,  so our already decaying function gets   a second modulation term e^-Rb times time. Rb  is the decay rate due to field inhomogeneity.   The nice thing about exponentials is that  the decay constants multiplying time just   add, and we can write a a decay rate  R* comprehensive of all sources of   spin dephasing. R* by this model is clearly just  equal to 1/T2 plus R sub B which we can write as   gamma times delta B which is just delta omega.  Similar to T2, R* can be characterized by its   own time constant which we designate T2* which  is the observed signal decay due to all the   dephasing mechanisms present in our sample. This  is problematic for us, because now our signal   decay is more representative of our magnetic  enviroment, not molecular properties intrinsic   to tissue. Thus our ability to contrast  tissue in an image is drastically reduced. To illustrate this, consider  3 tissues of various T2's and   equal Boltzmann magnetizations. Here are their  T2 decau plots in a perfectly homogeneous field,   delta b = 0. At an echo time of say 40  ms we can contrast the tissue very well. If our field inhomogeneity increases to just 1  part per million, our observed decay time T2*   is reduced to milliseconds. Not only do we have  to acquire our signals faster, but our ability   to contrast tissue is dramatically reduced, if  not gone depending on our signal to noise ratio.   Best case scenario is we'd be imaging a  tissue with infinite T2 - that would drop   this term out - but notice we can't  do better than 1 over gamma delta B.   And this part per million example I keep using  does indeed reflect most commercial MR scanner   limits. Manufacturers boast field inhomogeneity of  about 1ppm within the volume of the scanner bore.   Of course field homogeneity gets better  as you image a smaller and smaller volume.   Fields can be 10 to 100 parts per billion if  your field of view is down to 10s of centimeters,   which is quite impressive, but of course is  a large percentage of a scanner's price tag. But, it's important to remember that even if  you can manufacture a magnet with absolutely   perfect field homogeneity, Once you put  something to image inside - especially   living, physiologically active tissue - the  field will be distorted and delta B will   increase. So you can only go so  far perfecting a field anyways. Oh and you might have noticed that somewhere  along our discussion, we made a jump from having a   frequency distribution describing  our field inhomogeneity to a single   number delta B. What happened there? Shouldn't  it be, idk, some sort of integral or something?   Well, it so happens that if we model our  distribution using the Lorentzian lineshape   I mentioned earlier, delta B can in fact be  represented by a single number. This number   is most often related to the Lorentzian's full  width at half maximum (or half width at half   maximum depending on which textbook you read).  In our Fourier Theory video, we'll develop the   relevant math, but for now rest assured this  is a common, mathematically sound practice. Ok quick recap: the ping NMR experiment  involves exciting nuclear magnetization into   the transverse plane where its transverse  decay reflects our detected signaland   intrinsic properties of the tissue. But, because  our field is imperfect, this decay is accelerated,   and our ability to contrast tissue is reduced.  We're about to discuss a revolutionary method   for reversing the effect of field inhomogeneity  on our FID. But before we do, can you think of a   way get our spins back in phase after dephasing  due to field inhomogeneity? Perhaps our cartoon   of the color spins dephasing might help? Pause the  video if you'd like to ponder your own solution. Well... Consider the NMR experiment.   I'll draw our signal plot here and I'll use these  pink bars to indicate excitation pulses. Of course   the 90 degree pulse knocks our magnetization out  of the longitudinal plane and into the transverse   plane so we can detect the FID. A simple,  yet clever idea was suggested by Felix Bloch,   but first experimentally verified by  Erwin Hahn: just deliver a second pulse.   We will deliver this pulse soon after the FID has  decayed, and we'll make it a 180 degree pulse as   was suggested by Carr. Can you predict  what might happen to our detected signal?   Here's our 3D magnetization  vector cartoon to help. As expected, the FID is seen following the  90, but the addition of the 180 results in   a signal echo of sorts. This is called  the spin echo. So, what just happened? Well let's return to our dephasing spin cartoon  to help. The B0 field is coming out of the screen   here and I'll add our signal plot for us to  follow. Remember that the spin colors represent   their associated external field strengths:  red is in a high field, purple in a low field.   Notice what the 180 pulse does to our  dephased spins. After they're rotated 180   degrees their precession continue s, but  they eventually rephase with each other!   We can deliver another 180 if you'd like - the  result is another echo. Notice that after the 180,   the spins continue to precess at a rate dictated  by their field strength and represented by their   color: purples in low field are always precessing  counterclockwise, reds always clockwise. The 180   pulse doesn't change where the spins are in space  and so does not change their precession frequency. Perhaps this is easier to see if we look at the  animation of the spins unstacked. If you want   you can focus on an individual spin to verify  that its precession frequency does not change   following excitation. The spins dephase, then  following the 180 they rephase into a peak and   then continue to dephase. There's two important  things to notice about the plot I've drawn here:   first, you might notice that the dephase/rephase  parts of the plot seem to be mirrored   around the 180 pulses. Second, you see that our  echo peaks get smaller in time. What's going on? So as we've seen the detected FID follows  T2* decay. If we deliver out 180 at time tau,   then in that time the spins will  have dephased by a certain amount.   We should therefore expect that following the 180  the spins will take exactly that amount of time to   rephase. The echo peaks decay because unlike  dephasing due to field inhomogeneity, intrinsic   T2 decay is not reversible. Spins are still  losing coherence due to their self-interactions.   Here I'll draw the plot of T2 decay. Notice  that our peak comes right up to the T2 plot.   And in fact, this is the take-home.  If our 180 is delivered at time tau,   the echo will appear at time 2 tau and its  peak will reach the T2 decay curve. Thus the   peak amplitude can be used to calculate T2! We've  regained our T2 contrast. And this will be true Notice by the way that while I animated the  B1 pulse on the rotating frame's x axis,   echoes will appear regardless of the  excitation axis. Here's an example   with B1 at 45 degrees to x. We still get  an echo, but it appears along the y axis.   Thus we can expect our echo to  have a different phase as our FID. I'm going to simplify the animation here  with just a few spins to illustrate the   effect of delivering B1 at different  orientations in the rotating frame.   Notice the green spin does not precess here:  his delta B is zero, he's perfectly at larmor   resonance. As such we know all his buddies  will aligned with him at each echo. We can use this fact to help with our discussion.  Here's my dephased spins immediately before a 180.   Notice the angle our green spin  makes with respect to the B1 field.   Follwing the 180, our spin will  be at this angle on the other side   of B1. Thus we can expect our echo to  be at twice the phase of our B1 pulse. Ok so why does this work? Well the reason is that our spins have acquired   a certain amount of phase at time tau following  the 90 degree pulse. Let's examine a spin with a   negative gamma here so The angle theta of our spin  would just be omega t. In the argand plane, the   real component mu x reflects cos omega t and the  imaginary component reflects sin omega t. These   components can be visualized on a single plot and  imply a counterclockwise progression in phase. After a 180 degree pulse the imaginary component  becomes negative - but the real component is   unchanged. This takes our sin omega t and makes  it negative. The reversal of the imaginary   component mathematically switches the phase  we've accumulated from positive to negative:   minus theta. Our system will continue to evolve  progressing in phase counterclockwise at rate   omega t. The phase total accumulated phase  with thus be 0 when our time t equals tau,   and we'll detect and echo. Pretty neat huh?   The math is nice and simple here where  B1 is aligned along the real axis, but   perhaps you can extend the math for cases of B1  delivered at an arbitrary angle to the system... Oh and here's a leftover spin echo animation  in the laboratory frame just so you can   visualize what's going on. I only did  one of these since it makes me dizzy. So there you have it, the spin  echo. While we've focused on   dephasing due to field inhomogeneity as  the being reversible by the spin echo,   it will also reverse dephasing due to chemical  shift and heteronuclear spin interactions.   Spin echo sequences in MRI do have drawbacks which  we'll look at in later lectures. For one, they   take longer since echo times are increased.  And they involve much more rf delivery compared   with simpler gradient echo sequences. But  all of these details are for another time. And now you know the story of  how helpless NMR signals are   freed from the prisons of inhomogeneous magnetic  fields by the humble 180 degree refocusing pulse,   thus allowing for true T2 contrast in tissue. If you really want to put your newfound knowledge  to the test, try a few of these exercises.   Post your solutions in comments. See if  others got the same answers. Have fun! That concludes Part 2 of our little  MRI intro. Thanks for watching! And   Sorry it took so long to deliver. I do intend to continue this little intro series.  The next video will be all about Fourier Theory   and the Fourier Transform. Fourier theory is so  ubiquitous that another YouTube lecture may just   be a drop in the ocean, but it's so fundamental to  studying MRI that I can't NOT make one for the MRI   fundamentals series. I'm thinking 4 videos  total in this intro series with the fourth   being a discussion of how gradients produce  images and the Gradient echo pulse sequence.   After that I think we'd be well positioned to  discuss MRI topics in detail without leaving   anyone behind - things like pulse sequences,  MRI hardware such as spectrometers and coils,   chemical shifts, hyperpolarization, clinical MRI,  reconstruction and acceleration techniques, etc.   Maybe you have a question or topic  you'd like to see? Post a comment!   I'm definitely not a full-time YouTube  person so I simply can't respond to every   comment like some of the greats do, but  if I do see questions, and if they have a   lot of thumbs up's that'll let me know what  y'all would like to see in a future video. Cheers, friends.