Basic Principles of Magnetic

Resonance Imaging
Robert-Jan M. van Geuns, Piotr A. Wielopolski, Hein G. de Bruin,
Benno J. Rensing, Peter M.A. van Ooijen, Marc Hulshoff,
Matthijs Oudkerk, and Pim J. de Feyter

quently used nucleus, but other possible nuclei

Magnetic resonance imaging (MRI) is a noninvasive
imaging technique that is becoming more and more are carbon-13, sodium, and phosphorus. Under
important in clinical cardiology. Physicians must normal circumstances these tiny magnets are
understand the basic principles of MRI before reli- randomly distributed in space, the magnetic mo-
able use in practice is possible. Therefore, we will ments cancel each other out, and thus the net
give an introduction to basic MRI principles neces- magnetic vector is zero (Fig 1A). However, when
sary to understand the difficulties of cardiac MRI. the patient is submitted to a strong external
First the generation of a signal by the combination of magnetic field (B0) the nuclei adopt one of two
a strong magnetic field, radiofrequency pulses, and possible orientations: parallel or antiparallel to
temporary changes in the magnetic field is ex- the external field (Fig 1B). Parallel alignment is
plained. Then, the processes of localization of differ-
the lower energy state and is thus the preferred
ent points in an image, resolution, and signal-to-
alignment, whereas antiparallel alignment is the
noise ratio are highlighted. Finally, the influence of
tissue characteristics such as T1 and T2 on the higher energy state. The energy difference be-
contrast of an image are discussed. tween the two states is very small: the population
Copyright 1999 by W.B. Saunders Company ratio is approximately 100,000 to 100,006. A net
magnetization vector (Mz) aligned to the external
magnet results from the difference between the
two populations.
M agnetic Resonance (MR) imaging is an excit-
ing diagnostic imaging tool that uses strong
magnets and low-energy radiofrequency signals
Individual nuclei do not actually line up with
the magnetic field but wobble or precess around
(such as those found in radios and televisions) to the direction of the external field (Fig 2A). The
gather information from certain atomic nuclei frequency of this precession is given by the
within the body. Therefore, MR does not require Larmor equation:
ionizing radiation to obtain images. Several text- F B0 /2
books on magnetic resonance imaging (MRI) give
where F is the precessional frequency, B0 is the
excellent overviews of the basic concepts of MR
strength of magnetic field, and is the gyromag-
physics.1-4 Therefore, in this paper we restrict
netic ratio of the nucleus.
ourselves to summarizing these basic physics
This frequency is also called the Larmor fre-
concepts and refer the reader to the textbooks for
quency. In the frequently used commercial sys-
a more detailed description.

The Source of the MR Signal From the Department of Cardiology, Thoraxcenter, and
the Department of Radiology, Dr Daniel den Hoedkliniek,
A correct description of what happens when University Hospital Rotterdam, Rotterdam, The Nether-
tissue is subjected to a magnetic field relies on lands.
quantum mechanics. Fortunately, all the theory Address reprint requests to Robert-Jan M. van Geuns, MD,
Department of Cardiology, Thoraxcenter, BD 377, University
necessary for MRI can be based on a simple
Hospital Rotterdam, PO Box 2040, 3000 CA Rotterdam, The
classical model in which certain nuclei that spin Netherlands; e-mail: [email protected].
around their own axes behave like small magnets. Copyright 1999 by W.B. Saunders Company
For clinical imaging, hydrogen is the most fre- 0033-0620/99/4202-0003$10.00/0

Progress in Cardiovascular Diseases, Vol. 42, No. 2 (September/October), 1999: pp 149-156 149
150 VAN GEUNS ET AL

Fig 1. (A) Without a mag-

netic field the magnetic mo-
ments of the nuclei are dis-
tributed at random and thus
the net magnetization factor
is zero. (B) When there is a
strong external magnetic
field the spinning nuclei align
parallel or antiparallel to the
external field (B0) with a few
more parallel than antiparal-
lel. This results in a net mag-
netization vector (Mz) paral-
lel to the external magnetic
field.

tems of 1.5 Tesla (T), the Larmor frequency will patient, two phenomena occur: first, enough
be 63.75 MHz for hydrogen. It is of note that the protons absorb energy to jump from the parallel
phase of precession around the axis of the magnetic state to the higher level of the antiparallel state,
field is different for each individual nucleus (Fig 2B). and second, the spins are whipped to precess in
phase. The effect of all this is that the net
magnetization (Mz) flips 90 from the positive
Excitation
z-axis to transverse plane (Fig 3). The net magne-
The net magnetization vector from the nuclei tization in the transverse plane rotates around B0
inside the magnet in its equilibrium state is static at the Larmor frequency. This rotating transverse
and does not produce a measurable signal. To magnetization can be measured, because it will
obtain information from the spins, the direction induce an alternating current (AC) in the receiver
of the net magnetization vector has to be altered. coil placed around the patient.
For this the precessing spins are excited by
applying energy, in the form of radiofrequency
Return to Equilibrium
(RF) energy pulses of exactly the Larmor fre-
quency (resonance frequency). When an RF sig- After the RF frequency transmitter is switched off,
nal is given at the resonance frequency into the the equilibrium state will be sought (high energy

Fig 2. (A) In more detail the

individual nuclei spin around
their own axes and wobble or
precess around the direction
of the external field (B0). (B)
The phase of the precession
around the axis of the external
magnetic field is for each indi-
vidual nucleus.
BASIC PRINCIPLES OF MRI 151

Fig 3. When the spins are

exited with an RF pulse of
exactly the Larmor fre-
quency, the net magnetiza-
tion flips 90 and the spins
are whipped to precess in
phase. The rotating net mag-
netization vector induces an
AC in a receiver coil.

back to low energy). This means that the magneti- ation and longitudinal relaxation. Both processes
zation decays over time, which is represented by a are independent. The process of realignment to
decreasing magnitude of Mz in the transverse the external magnetic field is called the longitudi-
plane. Consequently, the induced signal in the nal relaxation process (Fig 5). It is characterized
receiver coil will decrease in time. This decreasing by the T1, relaxation time. The T1 relaxation time
signal is called the free induction decay (FID) is defined as the time required for the system to
(Fig 4). The time required for the signal to return recover to 63% of its equilibrium value after it has
to equilibrium is the relaxation time. been exposed to a 90 RF pulse (Fig 6). Various
Two relaxation processes exist: transverse relax- human tissues have different T1 values (Table 1).
The second process of relaxation, the trans-
verse relaxation, depends on the spins precessing
around the magnetization vector. Initially, after
the excitation by the RF pulse, the spins precess
completely in phase. However, as time passes, the
observed signal starts to decrease because the
spins begin to dephase due to small differences in
the Larmor frequency induced by random local
magnetic inhomogeneities, due to spin-spin inter-
action and inhomogeneity of the main static
magnetic field B0. This process is called the transverse
relaxation or spin-spin relaxation and is character-
ized by T2 relaxation time (Fig 5). The T2 relaxation
time is the time it takes for dephasing to decay the
signal to 37% of its original value (Fig 6). The T2
time from various tissues is different, but the T2
time is always shorter than the T1 time.

Spatial Encoding
Fig 4. The received signal detected by the receiver
coil, the FID, decreases over time when the net magne- To create an image, the MR signal from the
tization vector returns to its original orientation. H-protons has to contain information about where
152 VAN GEUNS ET AL

Fig 5. During relaxation, two

processes exist: longitudi-
nal relaxation and transverse
relaxation. Longitudinal re-
laxation (upper row) is the
realignment of the net mag-
netization to the external
magnetic field. Transverse
relaxation is the dephasing
of the precessing spins
(lower row).

these H-protons are positioned in the patient. combinations of gradients in all three directions,
This is done in three steps: slice selection, fre- it is possible to acquire a slice in any arbitrary
quency encoding, and phase encoding. direction through the body.
To select an imaging slice through the body, a The frequency and phase encoding are used to
magnetic gradient (for example 25 milliT/m) is obtain information for the individual points within
added along the main magnetic field in the caudal a slice, the picture elements or pixels. For the
to cranial direction. Because the frequency of phase encoding, a short temporary change in the
precession, and thus the frequency at which the magnetic field is applied between the RF excita-
spins can be excited, is dependent on the local tion pulse and the readout of the signal. This
strength of the magnetic field, a narrow band of change in the magnetic field will influence the
frequencies will only excite a thin slice (3 to 8 frequency of precessing, resulting in a shift in the
mm) of spins through the body (Fig 7). With a phase of precessing of the spins dependent on the
change in the excitation frequency another paral- duration of this gradient switch. By repeating this
lel slice can be acquired later. To obtain slices in process with different duration of the temporary
other directions, for example vertical slices, the gradients, signals with a different phase encoding
direction of gradients for the slice encoding are are acquired (Fig 8).
altered to an anterior-posterior gradient. By using The frequency encoding is used to differentiate
pixels with the same phase encoding. A magnetic
gradient during readout of the signal results in a
specific shift of the resonance frequency, likewise
the effect of the slice-encoding gradient, for pixels
with the same phase shift (Fig 9).

Table 1. T1 and T2 value of different tissues

at 1.5T
Tissue T1 (ms) T2 (ms)

Skeletal muscle 870 47

Fig 6. Longitudinal relaxation is characterized by the Myocardium 600 40
T1 relaxation time, which is the time to recover 63% of Liver 490 43
the original net magnetization vector. Transverse relax- Fat 260 84
ation is characterized by the T2 time, which is the time Blood 1,210 35
it takes to decay the signal to 37% of the original Venous arterial blood 1,210 250
signal.
BASIC PRINCIPLES OF MRI 153

Fig 7. A single slice through the body is selected in

MRI by superimposing a small magnetic gradient on Fig 9. Frequency encoding, with a gradient during
the main magnetic field in cranial-caudal direction. A readout of the AC signal, is used to differentiate pixels
RF pulse will only exite the spins, where the RF pulse with the same phase encoding.
matches the local precessing frequency determined
by the Larmor equation.

(Fig 10). The amplitude of each frequency repre-

Combining phase and frequency information sents the intensity of each pixel. The Fourier
allows the creation of a grid in which each pixel Transform is performed in both the frequency and
has a defined combination of phase and frequency phase encoding direction. Important to realize is
codes (Fig 9). This grid of raw data is called the that the imaging time for a single image depends
K-space. With a Fast Fourier Transform, the raw on the number of image lines desired, which is
data, which represent an amplitude as a function directly related to the number of signals with
of time, are transformed into a curve that repre- different phase shift that have to be acquired. For
sents an amplitude as function of the frequency example, for an image of 256 256 pixels, or 256
image lines, 256 signals have to be acquired.

The Echo Signal, Spin-Echo Imaging

There are several reasons why the FID signal is
not used for clinical imaging. First there is a
certain time necessary to perform the spatial
encoding, and even with present ultra-fast MR
scanners this can not be performed before the FID
declines. Second, the creation of a second AC
signal gives opportunities to modify the contrast
in the images depending on the T1 and T2 values
of the tissues.
To evoke a second AC signal, a second RF pulse
Fig 8. During phase encoding a temporary gradient is
applied. The change in the magnetic field will influence is applied that flips the spin by 180, and also
the precessing frequency. After the gradient is switched reverses the dephasing process (Fig 11). As the
off, the spins will precess with the original frequency, spins rephase, the amplitude of the AC signal
but a small change in the phase of precessing will
remain. The process has to be repeated to acquire increases and this signal, called the echo signal, is
multiple AC signals with different phase encodings. measured at its maximum (time of echo TE).
154 VAN GEUNS ET AL

Fig 10. A hypothetical AC

signal of a single image with
two pixels with a different pro-
ton density will result AC sig-
nal echo, with interference
pattern of 2 sinusoidal AC cur-
rents. A Fast Fourier Trans-
form extracts the different fre-
quencies and their amplitude
from the echo signal. The am-
plitude of each frequency rep-
resents the proton density in
each pixel.

Fig 11. Due to dephasing,

the FID signal rapidly de-
creases before the longitudi-
nal magnetization returns to
zero. A second RF pulses
(RF2), which flips the spins
by 180, also reverses the
dephasing process. When
the spins are in phase again,
a second AC signal, the echo
signal, arises. The interval
between the second RF pulse
and the echo signal is called
the time of echo (TE).

Fig 12. Contrast between different tissues. Two tissues each with a different T1 and T2 value have different signal
intensity dependent on the time. The contribution of T1 to the signal intensity of a pixel is dependent on the TR used,
indicated in the figure for a short and a long TR. From these points, the signal intensity decays along the T2 curve.
Using a short TR and TE, the signal intensity and contrast will be mainly influenced by the differences in T1 value of
the tissues (T1 weighted imaging). A combination of a long TR and a short TE will result in contrast dependent on the
proton density, whereas a long TR and a long TE will pronounce difference in T2 value (T2 weighted imaging).
BASIC PRINCIPLES OF MRI 155

mainly influenced by the difference in T1 value of

the tissues (Fig 12). Using a long TR and a long
TE, the contrast will be dependent on T2 differ-
ences. A combination of a long TR with a short TE
will not be dependent on T1 or T2, but only on the
proton density of the tissue.

Resolution
In digitized imaging, such as MRI, pictures are
composed of a matrix of elements, called picture
elements or pixels. The image represents the field
of view (FoV). The image matrix defines the
Fig 13. In MR imaging, resolution is determined by the number of pixels used to construct an image that
pixel size, which is dependent on the number of lines is determined by the number of frequency encod-
in the x and y direction used in the matrix and the FoV
covered with this matrix. Pixel size FoV (x)/No. lines ings (128 or 256 on the x-axis) and the number of
(x) X FoV (y)/No. lines (y). phase-encoding steps used (128 or 256 on the
y-axis) for a certain FoV (Fig 13). Therefore, the
MR techniques using the combination of a 90 FoV, the matrix size used, and the slice thickness
and a 180 RF pulse to generate an echo signal are determine the volume of each pixel.
called spin-echo sequences. In practice the resolution is determined by the
pixel size (the smaller, the higher the resolution),
but the signal-to-noise ratio is the limiting factor
Contrast if the pixels become to small and do not contain
MRI has the potential to visualize the difference enough spinning protons to produce a measur-
in T1 and T2 of different tissues. Using these able signal.
differences, contrast between different soft tissues
in MRI is superb compared with x-ray computer
Summary
tomography. If the time for the next repetition of
RF pulses (time of repetition TR) is shorter To obtain a single MR image, a complex combina-
than the time necessary for total longitudinal tion of RF pulses and magnetic gradient switches
relaxation, the contrast in the image will be have to be applied. In spin-echo imaging, two RF

Fig 14. Example of a spin-echo sequence. An echo signal is evoked after a 90 and a 180 RF pulse. Spatial
encoding of the echo signal is performed with gradient switches in three directions. The slice selection gradient is
present during RF excitation, which will now only excite a thin slice through the body. The phase-encoding gradient
is shortly present between the excitation and the sampling of the echo signal. Finally, the frequency-encoding
gradient is activated during the echo acquisition. This combination of RF pulses and gradient switches is repeated
128 or 256 heartbeats with different phase-encoding steps. Imaging time will be approximately 3 to 4 minutes for a
single cardiac triggered image.
156 VAN GEUNS ET AL

pulses combined with gradient switches in three References

directions result in one echo signal with spatial
information. For an electrocardiogram-triggered
cardiac image, in each heartbeat one combination 1. Stark D, Bradly WG: Magnetic Resonance Imaging. St
is completed and thus one echo is acquired (Fig Louis, MO, Mosby Year Book, 1992
2. Rinck P: Magnetic Resonance in Medicine. Oxford, UK,
14). To complete a single image of 256 lines, 256
Blackwell Scientific Publications, 1993
echos with different phase encodings need to be 3. Heiken JP, Brown JJ: Manual of Clinical Magnetic
sampled; therefore, imaging time, for a single Resonance Imaging. New York, NY, Raven Press, 1991
spin-echo image, will be 256 heartbeats, which 4. Rayan SS: MRI: A Conceptual Overview. Berlin, Ger-
takes in most patients between 3 and 4 minutes. many, Springer, 1997