1. My spin on MRI: The basics of MRI physics and image formation.
Jonathan Dyke, Ph.D.
Assistant Research Professor of Physics in Radiology
Citigroup Biomedical Imaging Center
Weill Cornell Medical College
Sackler Institute for Developmental Psychobiology
Summer Lecture Series
July 7, 2009
Good morning and I’d especially like to thank BJ for inviting me to speak on these various imaging modalities.
The work that is done here at Sackler along with many projects that we’re involved in at the Citigroup Biomedical Imaging Center focus around state of the art neuroimaging techniques.
I’m going to give a brief overview of Magnetic Resonance Imaging, it’s basic principles and applications to the work done here in the remainder of this session.
I’ll follow up on Thursday by introducing additional imaging modalities such as Positron Emission Tomography and illustrating how they may complement and add to the knowledge gained via MRI.

2. Now not all nuclei are “MRI active”..
Which of the following could produce an MRI image?
Only those with an odd number of protons and neutrons. 1H
11C
13N
18F
19F
Spin ½ particles having an unpaired spin due to an odd number of protons and neutrons can produce a magnetic moment detectable by MRI.
The spin states would then be spin-up and spin-down and preferentially align in a specific direction in a strong magnetic field.
Even nuclei or paired particles cancel each other out and do not possess an MR excitable spin.
I purposefully put the Positron Emission Tomography tracers in the list to tweak some interest for hanging around until Thursday.
Some of these isotopes are radioactive and have specific half-lives of decay.. by causing tissue changes and damage.
31P
Which isotopes at the right are radioactive?

3. Does an MRI scanner produce radiation?
Does an MRI scanner produce radioactivity.. The energy difference between the spin-up and spin-down states produces a photon with an energy in the radio frequency range that does NOT produce ionizing radiation. The MRI nuclides in use are not radioactive and no ionizing radiation is produced in an MRI scanner..
What exactly is ionizing radiation? Ionizing radiation is capable of detaching electrons from atoms thereby producing damage.
What detrimental biophysical effects are produced by an MRI scanner? - babies pointing North - joke?
Static magnetic field – pacemakers, clips, shrapnel, projectiles ; changing field – dB/dT,
The MRI signal is generated by receiving radiofrequency
photons that return to their lower energy state.

4. of the positively charged _______.
A hydrogen atom (whether bound in water or lipid) acts as a small magnet due to the spinning
of the positively charged _______.
What 3D chemical structures are displayed above?
The MRI signal that produces the images we see comes from protons in the body primarily present in water and lipid molecules.
Also, note that it is the spinning proton (as we’ll show in a minute) that is responsible for the dipole moment that is detectable in MRI.
proton

5. Protons from what compounds comprise an MRI signal?
What percentage of your body is composed of water?
What percentage of your body is composed of fat?
A) 40%-50%, B) 50%-60%, C) 60%-70%, D) 70%-80%
Description
Women
Men
Essential fat
10–12%
2–4%
Athletes
14–20%
6–13%
Fitness
21–24%
14–17%
Acceptable
25–31%
18–25%
Overweight
32-41%
26-37%
Obese
42%+
38%+
I’d also like this to be an interactive and informal session such that you can feel free to stop and ask questions of me when you want to dig in a bit further on a topic.
I’ll also be doing the same and will try to get your minds stirring and focused on the imaging science at hand as we go along.
Also, we’re going to keep the equations and math modeling a bit more subdued and make sure that you can qualitatively and visually recognize and explain various techniques offered by MRI.
The human body is 60-70% water.. As for the % fat.
For those guys out there, this is really a question that you should never bring up in the presence of any girl you wish to impress!

6. When placed in a “very” strong magnetic field,
a “slightly” greater number of atoms
align in the preferred direction with the magnetic field.
Vs.
Basic MRI texts that you may get your hands on use the above terminology to describe the first step in receiving an MRI signal. The patient is placed in the MARI scanner and a slightly greater # of atoms align with the main magnetic field.
A junkyard car crane magnet is actually only 1.0 Tesla or 10,000 Gauss.. All standard MRI scanners are 1.5 Tesla and we have (3) 3.0 Tesla scanners that are used on campus and primarily for all experiments at Sackler.. A small toy magnet is approx 100 gauss, a strong neodymium is 2000 gauss
The CBIC 3T MRI scanner is three times more powerful than A junkyard electromagnet that lifts cars from the yard.
“very”: The earth’s magnetic field is ~ 0.3 Gauss.
The scanner at our facility is 30,000 Gauss.
“slightly”: only 5:1,000,000 spins align!
Note: 1 Tesla = 10,000 Gauss (Metric)

7. Typical Magnetic Field Map of a Clinical 3T MRI
What effects will be felt by a pacemaker,
credit cards, earrings, IPAD or cell phone?
5 Gauss line or 0.5 mT is the FDA required limit for patients with pacemakers. Under absolutely no condition should a pacemaker be brought near an MRI scanner.. 1 mT – 2mT or Gauss are required to erase cards and electronic media.

8. The MRI scanner is always on!! A magnetic field is present 24/7!!
An aside here, as an MR physicist, there is truly no better way to illustrate the fact that patient screening and ferromagnetic detection is a life or death situation when working around high field magnets.
During your stay here at Sackler, I’m sure that you will get to visit an MRI Scanner and observe several imaging trials. The most common question is whether the magnet is always on. Unlike a junkyard magnet, these are superconducting, liquid helium cooled magnets that once energized remain on 24/7.
A healthy respect should be used for all scanners and I’ve seen several small incidents and close calls over the years that warrant my respect.

9. MRI Safety Implants and foreign bodies Projectile or missile effect
Radio frequency energy
Peripheral nerve stimulation (PNS)
Acoustic noise
Cryogens
Contrast agents
Pregnancy
Claustrophobia and discomfort
-“Cheap” Earrings
- Tattoo Ink
> Rock the gardens.
- “Quench”
- Nephrogenic Systemic Fibrosis
As an MR physicist I just want to mention the safety factors associated with MRI.. Specialized scans such as MR contrast administration also have considerations.
- No X-rays/Gd crosses placenta.

10. How does resonance come into play in MRI?
A typical tuning fork produces a frequency of 400 Hertz,
while a scan from Sackler was actually resonating at
127, 503, 172 Hertz.
Back to formation of an MRI image, we now have a preferential alignment of the magnetic dipoles present in the tissue with the main magnetic field. In order to receive a signal from these static dipoles, we perturb the system by rotating them a known angle away from the z-axis along the bore.
A tuning fork produces sound waves at a single frequency that may be detected by objects that are of lengths related to multiples of the wavelength.

11. w=Precessional Frequency g= Gyromagnetic Ratio
Larmor Equation: w=gB
w=Precessional Frequency
g= Gyromagnetic Ratio
B=Magnetic Field Strength
(42.57 MHz/Tesla * 3.0 Tesla = MHz)
What field strength does my favorite
FM Classic Rock station transmit at?
Radio waves are transmitted at an angle of 90˚
into the body at the Larmor frequency.
This imparts energy to the nuclei to achieve “resonance”
The additional energy in turn rotates the nuclei
out of alignment with the main field.
This is “THE” one main equation that you should remember from this talk on MRI.. Simple and succinct.
Note that the RF band in New York City is ripe with transmitting frequencies.. Wireless equipment, FAA flight plans, and much more transmit in the hundreds of megahertz that would produce visible interference on the images. The MRI suite must be radiofrequency shielded in a “faraday cage” by a copper lining and mesh in the glass to shut out all frequencies from outside the scanner room.. You can (and we do) take a transistor radio (BJ remembers those!) into the scan room to check for RF leaks.
Also note that the gyromagnetic ratio is characteristic of the nuclei that is being excited.. I also scan 31-phosphorus studies and my excitation frequency is only 51.5 Mhz to see various phosphomonoesters, phosphocreatine and ATP peaks in spectroscopy.

12. Y X Z 3.0 Tesla GE MRI Scanner Coil
Just a visual aid to show that the dipoles are preferentially aligned with the Z-axis on the MRI scanners.. A radiofrequency pulse of known amplitude and duration then “flips” the spins a known angle away from the Z-axis. Typically 90 degrees for a spin echo sequence into the X-Y plane but also very commonly lesser angles such as 12 degrees for 3D gradient echo sequence which we’ll discuss later.. Also 180 degrees is also commonly used for saturation and inversion recovery sequences.
As a caveat, we found out that the magnet at CBIC is actually wound opposite to those at NYP and WGC.. Our scanner was one of the first built and produced in England at Oxford magnet for General Electric and the British winding convention is actually opposite to the handedness of the scanners now being produced for GE in South Carolina.. So basically the North and South poles are flipped.. We know this because I’ve borrowed coils from the hospital and Doug figured out that they needed to be physically rotated on our scanner to work properly.

13. Was it easier back then to get a law named after you?
“Magneto”
So now that we have rotated the magnetic dipoles in a known direction, how do we record a signal from the tissue?? How do we receive an RF photon produced by the sum of millions of magnetic dipoles realigning with the main magnetic field produces a time varying magnetic flux.
This results in a changing voltage in a fixed RF coil which is placed near the specific area of interest in the study such as a head coil..
Faraday’s Law of Induction states that a voltage is created by a changing magnetic flux. (1831)
Was it easier back then to get a law named after you?

14. How do the motion of these two objects differ?
“Rotation” vs. “Precession”
Just to be proper, I’d just like to mention that in the laboratory frame of reference that the dipoles realign with the main magnetic field as a spinning top which is called “precession” and is accomplished in 3-dimensions. In the mathematical rotating frame of reference the spins perform a 2D rotation.
If the dipoles are rotated from their plane of origin, the orientation of the dipoles is followed by a precession of the magnetic moments as they try realign with the direction of the main magnetic field in the MRI scanner.
It is the precession of the nuclei that creates the
changing magnetic field needed to produce a signal.

15. What kind of signal is actually received by the scanner?
The frequency & phase information in time from the Free Induction Decay “FID” are transformed
into the frequency domain. (NMR 1946)
A Fourier series can represent any function
as a sum of sines and cosines. (1822)
So as the dipoles are precessing to realign with the main field they produce a time varying voltage that looks like a damped oscillation which resembles a classical spring system that you may have studied in undergrad physics. The maximum voltage received occurs when the spins are 90 degrees perpendicular to the z-axis in the X-Y plane of the RF coil. As they realign with the z-axis, the induced voltage in the coil then returns back to zero.
The signal shown above is called a free induction decay or “FID” and is the signal directly detected by the MRI coil after the spins realign. This is actually the result of a Nuclear Magnetic Resonance (NMR) scan that you may have encountered in Chemistry. The signal can be Fourier transformed in order to produce a frequency spectrum of the water, lipid and other metabolic peaks in the tissue of interest. Concentration wise, an MRI signal

16. Typical NMR signal after Fourier transformation.
Can you identify the peaks? How about concentration?
H2O (4.7ppm)
Lipids CH2 (1.3ppm)
As my personal research interests focus on Magnetic Resonance Spectroscopy and Perfusion weighted imaging, you’re going to get a slight NMR bent to this talk as well..  Notice that the water and lipid peaks are at different resonances or frequencies on the x-axis. The scale is known as parts per million and corresponds to a unitless scale that is converted to Hertz based upon the scanner frequency of origin.
As in 1D Chemical NMR, the area under the curve or the integral produces a measure of the concentration of the metabolite of interest. In standard anatomical and functional neuroimaging, we are looking at the area under the water peak which results in a concentration of around 80 Molar in-vivo. The signal intensity or brightness in an MRI image is then produced by summing the area under this curve.
As we move on, please focus on the area underneath the red box… not much visible there…
Lipids CH3 (0.9ppm)

17. Where were all of these metabolic peaks hiding?
This is an ex-vivo mouse brain perchloric extract that I had run off at 11.4 Tesla in the NMR core facility. The typical concentration of brain metabolites found in human subjects in on the order of 1 mM to 10 mM in comparison to 80 Molar.. There is then a difference in concentration of around 10,000 less in the metabolites compared to normal concentrations of water in the brain. Special water suppression and excitation pulses are used to detect these at lower spatial resolutions of approx 0.75 cm in volume compared to an imaging voxel of cc or (1mm x 1mm x 1mm).
So then why can’t we see all of these metabolites in a typical brain spectra? The frequency separation of the metabolites is directly related to field strength and the splitting is then increased with magnetic field. Typically at 3.0 Tesla you’ll see the CRE/CHO/NAA in healthy brain tissue and LAC in pathology. Each metabolite has a specific biological function in the human body and provides information on that pathway or cycle in use.
What price is paid in detecting these signals?

18. Damadian’s Design for a Clinical MRI Scanner - 1974
Let’s briefly take a look at how we can now acquire spatial information from an MRI scanner.. Or how do we acquire a FID from each pixel or voxel in an MRI image.. i.e. a basic 256 x 256 image has 65,536 pixels of which probably 2/3 contain imaging voxels within the brain. This is a lot of signal to process and a lot of information to encode.
Back when disco was on the rise, Damadian applied for a patent for the 1st clinical MRI scanner to produce non-invasive in-vivo images of a human subject. The first MRI image was acquired in 1977 and not until 1986 was the first scanner in the NYC area rolled into NYP. The delay of incorporating the 1st clinical scanner until 1986 really was awaiting the computer world to catch up as well as the onset of superconducting magnet technology.. Which is a monumental achievement in itself!

19. Basic MRI Hardware Block Diagram
Here is a block diagram of an MRI scanner.. A radio frequency electromagnetic field is briefly turned on, causing the protons to absorb some of its energy. When this field is turned off the protons release this energy at a resonance radio frequency which can be detected by the scanner. The frequency of the emitted signal depends on the strength of the magnetic field. The position of protons in the body can be determined by applying additional magnetic fields during the scan which allows an image of the body to be built up. These are created by turning gradients coils on and off which creates the knocking sounds heard during an MR scan.
Fast imaging sequences such as those used in functional neuroimaging can play upwards of 100+ decibels inside the bore of the scanner..
How many of you have had an MRI? What’s it like?

20. How loud is loud?
20 dB 30 dB 40 dB 50 dB 60 dB 70 dB 80 dB
Ticking watch Quiet whisper Refrigerator hum Rainfall Sewing machine Washing machine Alarm clock (two feet away)
85 dB 95 dB 100 dB 105 dB 110 dB 120 dB 130 dB
Average traffic MRI Blow dryer, subway train Power mower, chainsaw Screaming child Rock concert, thunderclap Jackhammer, jet plane (100 feet away)
Fast imaging sequences such as EPI/Spiral used in functional neuroimaging (fMRI) can play upwards of 100+ decibels inside the bore of the scanner.
As an aside, measurement of acoustic levels are made via the power decibel system which is logarithmic in nature.
About Twice as Loud 10dB , About Four Times as Loud 20dB
This is a serious concern especially for the Sackler lab in imaging pediatric subjects. Great care has to be taken to reduce the inherent scanner noise through earplugs and headphones while also allowing verbal instructions to be passed along to the subject.

21. So how do we get spatial information?
Back to the Larmor equation.. w=gB
Magnetic Field Strength
Position
We must then apply a linearly increasing gradient field in a specific direction in order to be able to assign a specific frequency to each voxel along that direction. Then the NMR peak of water will be slightly shifted to a different frequency for each voxel in the image. This can be translated back to a specific spatial shift.
i.e. 1 Gauss will increase the frequency by 4.3kHz.
Typical gradient strengths are 2-5 Gauss/cm.

22. What would the frequency difference be
between two objects that are separated by 3cm?
w=gB
= 42.57E6 Hz/Tesla
B = Gz * z
= 0.01 T/m * 0.03 m
w = 12,771 Hertz
freq(1cm) = g Gy y = 42.58E6 Hz/T * 0.01 T/m * 0.01 m = 4258 Hz
Freq (-2cm) = 42.58E6 Hz/T * 0.01 T/m * m = Hz

23. Conventional 3-Axis MRI Gradient Coil Diagram
An asset about MRI that is not available in CT or PET is the ability to prescribe oblique planes (or double oblique) of interest due to additive combinations of x, y & z gradients. Note that in order to properly obtain accurate serial imaging studies upon a subject that alignment of 2D imaging slices to the same landmarks should be accomplished.. For fMRI the alignment of the imaging plane to the soft palette or anterior and posterior commissure points is traditionally used.

24. 1st step is to excite a single slice instead of all space!
Slice Selection
1st step is to excite a single slice instead of all space!
Frequency
To excite a thickness z use: DZ=Dw/gGZ
To excite off axis use: w+w0 where w= gGZ DZ
In order to specifically excite a certain slab or slice thickness a combination of two things must occur:
A radiofrequency pulse must be transmitted with a fixed bandwidth that will only excite or resonate spins within a certain spatial range..
i.e. we found out previously that our example of a 3cm separation was on the order of 12 kHz.. Well for a 3 mm slice thickness a period of the
Transmit pulse may correspond to 1 Khz approximately.
2) A linearly increasing gradient in the slice direction (which for axial/transverse scans is the z-direction) must be applied at the same time as
The transmitted RF pulse in order to read in spins from only that slice…
How thin a slice depends on the gradient strength and speed of the scanner.. Typical slice thicknesses of 1mm are used clinically while 0.25 or 0.5mm are obtained in 3D acquisitions on the 3.0 Tesla scanner.. The 7T animal scanner that is being energized in our center today and tomorrow will reduce these numbers by at least two-fold.

25. How thin a slice could an MRI scanner produce?
Slice Selection
Shown here is an example of representative magnetic field and frequency shifts required to produce various axial slices in a patient.
How thin a slice could an MRI scanner produce?
i.e. Could we perform in-vivo pathology scans?

26. General Electric Spin Echo Pulse Sequence Diagram
Readout
Read
Rewinder
Phase Encode
Rewinder
Phase
Slice Select Gradients
Slice
TE/2
TE/2
Shown here is the actual spin echo imaging sequence from our clinical scanner as viewed in a pulse sequence simulator.
You can see that both radiofrequency and gradient pulses are applied to both excite and receive signal from specific regions in space.
90°
180° TR

27. Explaining the spin echo pulse sequence
Ready,
Set,
Go!!
Gun starts
With 90 deg pulse.
Courtesy: Siemens

28. Runners fan out with ability
The MRI correlate is that dephasing or spreading out of the spins occurs due to variances in the magnetic field as well as differences in relaxation times of each tissue.

29. Gun fires again reversing direction of race.
[180 deg pulse]

30. The runners then reach the finish line at the same time TE.
Maximum signal is reached when all the spins are aligned with the same phase in the receive coil at the same time.

31. General Electric Spin Echo Pulse Sequence Diagram
Readout
Read
Rewinder
Phase Encode
Rewinder
Phase
Slice Select Gradients
Slice
TE/2
TE/2
Shown here is the actual spin echo imaging sequence from our clinical scanner as viewed in a pulse sequence simulator.
You can see that both radiofrequency and gradient pulses are applied to both excite and receive signal from specific regions in space.
90°
180° TR

32. Now that we have selectively excited a specific slice
in space, we then must localize a specific xy-plane.
With what pattern is MRI data generally acquired?
Why would you choose one over the other?

33. Spatial encoding in x is called “Frequency Encoding”.
The frequency of the signal ~ position on the x-axis.
dx = FOVx/Nx = 1/(g/2p Gx tx)
e.g. A standard brain scan uses a 24 cm FOV
and a 512x512 matrix size on our 3T magnet.
This gives an in-plane resolution of 0.47mm/pixel.
RBW = Nx / tx = 1 /DT
e.g. A 15.63kHz RBW and Gx = 0.3 G/cm would
then apply the x-gradient for 32.8 ms to get
a single line of image “k-space”.

34. Spatial encoding in y is called “Phase Encoding”.
The phase f of the signal ~ position on the y-axis.
dy = FOVy/Npe = =1/(2 g/2p Gyr ty)
The phase of a signal is given by: f = w t
To acquire the next line in “k-space”, an additional
phase (gGyy) is applied for a time t.
This is repeated until the entire image space is covered.
It is standard for the time to be fixed and the
gradient amplitude to increase/decrease.

35. Why is a Fourier Transform used?
Application of pulses in the “time” domain
are transformed into the MRI “frequency” domain.
As an aside.. The RF Profile needed to excite a rectangular band of frequencies corresponds to a sinc pulse being generated by the scanner in real time.

36. FT K-space vs. Image Space FT

37. k-space Contribution to Image Properties
Center = contrast
Periphery =
resolution

38. Voila’ - Spin Echo Images

39. How does an MRI scanner differ from a CT scanner?
Radiation, 2) Soft-Tissue Contrast
The intensity on a CT scan is directly related to what?
How much energy does MRI impart?
EMRI=h(g/2p)B0 =0.3 meV vs. ECT~ 25keV CT T1 T2

40. T1W GM=950ms WM=600ms T2W GM=100ms WM=80ms
Image Weighting in MRI – * Learning Point *
T1W
GM=950ms
WM=600ms
T2W
GM=100ms
WM=80ms

41. Summary:
Magnetic Resonance Imaging
Soft Tissue Contrast (GM vs. WM, etc.)
High Spatial Resolution ( 1 mm isotropic voxels)
Oblique scanning options
Additional functionality:
Diffusion MRI, Magnetization Transfer MRI
Fluid attenuated inversion recovery (FLAIR)
Angiography, CSF Dynamics, Spectroscopy
Functional MRI, Interventional MRI, Contrast agents
MR guided focused ultrasound, Multinuclear imaging
Susceptibility weighted imaging (SWI)