Magnetic Resonance Imaging (MRI)

Nuclear Magnetic Resonance (NMR)

Nuclei in atoms have neutrons and protons. These subatomic particles each have spin. In nuclei with odd mass numbers these spins do not cancel each other out, hence will have a net angular momentum/spin (MR active nucleus). As per Faraday’s law of electromagnetic induction a spinning change will induce a magnetic moment. MR active nuclei include H, C, N, O, F, Na, P. In the human body hydrogen (H) is very abundant and because of its solitary proton has a relatively large magnetic moment.

Hydrogen nuclei, due to its magnetic moment will align parallel to an external magnetic field (spin-up) if it has lower energy, or anti-parallel (spin-down) if it is in a high energy state (Boltzmann’s distribution). There will usually be more protons spin-up, creating a net magnetisation vector (NMV). The stronger the B0, the more spin-up.

Precession = secondary spin/wobble around the external magnetic field (B0).
ω0 (Larmor/resonant/precession frequency) = γ × B0
B0 = external field strength (in T, tesla).
γ = gyromagnetic ratio (constant for a particular nucleus), which for hydrogen is 42.57 MHz/T


When an external RF pulse with energy the same as ω0 is applied perpendicular to B0 then resonance occurs. Nuclei gain energy from the RF, hence increasing the number of spin-down nuclei. The higher amplitude and duration, the more nuclei will change. The NMV is a vector, so has longitudinal (Mz) and vertical (Mxy) components. Initially the NMV is completely longitudinal; resonance will rotate the NMV, ie greater Mxy and less Mz magnetisation. With a flip angle of 90° the spin-up and spin-down populations are equal, cancelling each other out (hence no Mz), but with largest Mxy.

Nuclei will naturally precess randomly (out of phase/incoherent). Resonance will align these spins so they are all in phase/coherent. Since the spins are coherent with a transverse magnetisation, as per Faraday’s law a voltage will be induced into a receiver coil, producing MR signal (with frequency = ω0).

When the RF pulse is switched off relaxation occurs, with longitudinal plane of NMR increasing (T1 recovery) and transverse NMR reducing (T2 decay). This results in reduced signal in the coil = free induction decay (FID).

T1 Recovery

Spin-lattice relaxation – the nucleus gives up energy to surrounding molecules/lattice, increasing proportion of lower energy spin-up nuclei. Energy is transferred to a neighboring molecule when it is tumbling at the same frequency as the nucleus ω(similar to resonance). Recovery is exponential, with T1 relaxation time = time for longitudinal magnetisation to regrow to 63% (=1-e-1) of max value. T1 time reduces when there is less energy in the lattice (hence able to absorb energy from H).

  • Water (small molecules) have a fast molecular tumbling rate (which is not matched to ω0) and high intrinsic energy (less able to absorb), hence slow T1 relaxation.
  • Fat (medium-sized molecules) have relatively slow tumbling rate matching ω0 and less intrinsic energy, hence fast T1.
  • Large molecules (proteins, DNA, bone) tumble too slowly, so slow T1
  • Paramagnetic ions/molecules (transition group [Fe, Mn, Cu, Cr, Co, Ni, Tl], rare-earth [Gd], O2) have unpaired electron hence have a magnetic moment (700x H), creating a large fluctuating magnetic field with fast T1

T1 increases with B0 due to reduction in spectral overlap between molecular tumbling/vibration frequency and ω0

Fat regrows longitudinal magnetisation faster than water. If a second 90° RF pulse is applied (with TR < relaxation of tissues) then water NMV are flipped closer to 180° than fat; ie fat are closer to 90°. Hence fat will have high MR signal and water low signal.

T2 Decay

Spin-spin relaxation – neighboring nuclei interact with each other resulting in loss of coherency, ie nuclei will spin more out of phase. Decay is exponential, with T2 relaxation time = time for transverse magnetization to reduce to 37% (=e-1) of initial value. T2 decay is always ≤ T1 recovery. T2 time reduces when molecules are packed tighter together (more spin-spin interactions), inhomogeneity of local magnetic field (differences in ω0).

Magnetic field inhomogeneity is from external magnetic field imperfections (magnet construction), intrinsic fields (from neighbouring spins), or sample induced inhomogeneity from magnetic susceptibility/polarisation distorting local field (metallic objects, calcium, bone-air interfaces, haemosiderin)

T2* = decay of Mxy from all sources intrinsic and extrinsic

True T2 decay = from intrinsic fields only

  • Water – Molecules are spaced further apart. For pure liquids and gas there is no significant intrinsic static field. Hence long T2, more coherence, higher MR signal.
  • Macromolecules (fat, protein) – Molecules are packed closle together increasing chance of spin-spin interactions. Large intrinsic fields. Hence short T2, nuclei spins are more out of phase, less MR signal.

T2’ = decay from inhomogeneous extrinsic fields. Heterogeneity of the field results in different B0, hence differences in ω0 (=γ×B0), thus protons spins dephase faster. This T2′ decay is expontential, the predominate component of T2*.


Intrinsic contrast parameters (which are inherent to body tissues and cannot be changed) include:

  • T1 recovery time – as above
  • T2 decay time – as above
  • Proton density (PD) – number of mobile hydrogen nuclei per volume of tissue. Higher PD = more signal.
  • Flow
  • Apparent diffusion coeffecient (ADC).

Extrinsic contrast parameters include TR, TE, flip angle, TI, turbo factor/echo train length, b-value.

Fat has short T1 and short T2 time. Water has long T1 and T2.

tissue T1 (ms) T2 (ms)
water 2500 2500
CSF 2000 300
fat 200 100
liver 400 50
grey matter 800 100
white matter 700 90


Weighting uses extrinsic contrast parameters to control relative contribution of intrinsic parameters in an image.

  • T1 weighting – TR must be short so tissues do not have enough time to fully T1 recovery. Differences in T1 times between tissues are then demonstrated (contrast). TE is short so T2 is minimised.
  • T2 weighting – TE must be long enough for tissues to decay, so emphasize differences in T2 decay times between tissues. TR is long so T1 minimised.
  • PD weighting – Always present to some extent. T1 and T2 effects must be diminished, hence long TR (T1 fully recovered) and short TE (not enough time for T2 decay).

Partial saturation occurs when NMV >90° (resulting in T1 weighting). Full saturation occurs at 180°.

Pulse Sequences

T2′ effects cause rapid loss of

RF pulse = RF_pulse1 =RF_pulse2

Hahn echo = FID detected = FID1=FID2

upward/positive or downward/negative gradient = Positive_gradientor Negative_gradient

phase encoding gradient =Phase_encoding_gradient

TR = repetition time (ms) = time between RF pulses. Determines the amount of longitudinal relaxation that occurs between pulses. Short TR usually <500ms, long TR >1500ms.

TE = echo time (ms) = echo delay = time between RF pulse and peak of received signal. Determines the amount of transverse decay which has occurred. Always shorter than TR. Short T@ usually <30ms, long TE >90ms

Spin-Echo (SE)


to remove effect of H-field inhomogeneity

1/T2* = 1/T2 + 1/T2’ = 1/T2 + γ.ΔH0; T2’ = relaxation from local static field inhomogeneities; ΔH0 = constant for a magnet

90° pulse gives T2* echo; 180° pulse at τ = TE/2 flips the partly dephased spins into mirror image positions so that after rephrasing T2’ is removed

T2 echo amplitude is decreased relative to T2* echo by e-TE/T2 due to true T2 dephasing

amplitude/image brightness instantaneous value of Mxy

pulse sequence repeated at TR, rare for TR to be long enough for Mz to recover

T1 weighting: short TR (700ms, contrast develops during early regrowth) and short TE (10-20ms, to reduce T2)

PD (proton density) weighting: long TR (2500ms,for Mz to restore) and short TE (10-20ms)

T2 weighting: long TR (2500ms), long TE (80ms)

single-echo SE = single 180° pulse per 90° pulse

multiple-echo SE = equally spaced 180°, PD used from 1st echo (TE1=15ms), T2 from 2nd echo (TE2=100ms)

turbo/fast SE = echo-train SE = different phase encoding gradients applied between each 180° pulse until signal intensity is lost from true T2 exponential decay; each subsequent echo has lower SNR; best used in rectangular FOVs (spine)

Inversion Recovery (IR)


to emphasize differences in T1, longer 180° pulse used to emphasize the differences

first inverted by 180° pulse, gradual T1 recovery then 90° pulse (to flip protons that had already aligned, into transverse plane)

by RF rotating cone of spin up/down theory, after 180° pulse there is reduction in size of longitudinal magnetisation so when cone rotates at the 90° pulse there is little transverse magnetisation

TI = inversion time = time betw 180° and 90° pulses

additional 180° pulse is given after 90° pulse (ie SE after IR) and remains T1w with short TE (10-20ms)

previously TI needed to be >T1 for positive Mz (hundreds of ms) hence long scan times, but now with quadrature coils TI can be < T1 with Mz negative

FLAIR = fluid attenuated IR = TI is ≈70% T1 of CSF, hence CSF suppressed and good contrast with lesions adjacent to CSF/ventricles easier to see

STIR = short TI IR = TI is ≈ 70% T1 of fat, hence fat signal suppressed, long T1 tissues (fluid) will be bright

T1 weighted: long TR (2000ms) and short TI (400ms)

PD weighting: long TR (2000ms) and long TI (700ms)

Double inversion recovery (DIR). Reduced motor strip cortex signal compared to other gyri.

Calculated T1 and T2 Values

T2 = TE/ln(I1/I2) with measurements of intensities from two echoes that can be obtained in multiple-echo SE

T1 can be calculated from two different SE with different TR but same TE (with motion causing error); or from modified IR

Gradient Recalled Echo (GRE)


low flip angle <90° (α) pulse given, means low values of TR possible (10ms), but cannot use 180° rephrasing pulse and loss of T1 weighting

a negative gradient dephasing lobe on frequency encoding axis precedes the positive readout gradient to cause dephasing then rephrasing in the first half of the positive gradient; allows for very short TE (5-10ms)

does not recover dephasing due to inhomogeneities in H0, hence signal a function of T2* and T2; most marked at interfaces, enhancing magnetic susceptibility (haemorrhage [Fe], Ca, Gd)

since TR < T2 spoiler pulse may (incoherent) or may not (coherent = refocused) be given along Gz to dampen remnant transverse magnetisation before next cycle and create a steady state

Functional MRI (fMRI)

uses BOLD (blood oxygenation level-dependent) where deoxyhaemoglobin is paramagnetic shortening T2 and T2*

neuronal activity causes localised increase in blood flow, hence reduced deoxyhaemoglobin thus longer T2/T2* and higher signal; requires rapid sequences

Echo Planar Imaging (EPI)


a 90° pulse with slice selection performed

GRE EPI: the readout gradient is switched rapidly between negative and positive values (defocusing and refocusing), with phase encoding gradient performed between each switch to create multiple FIDs in quick succession

SE EPI:switching defocusing/refocusing gradients given after a SE sequence

GRASE = combination of SE and GRE

single shot EPI = an entire k-space is acquired in each sequence

multishot EPI = partial set of echoes acquired

total readout time must be short compared to T2/T2 *, or non-linear artefacts in phase encoding direction introduced

more vulnerable to magnetic susceptibility than GRE

Diffusion-Weighted Imaging (DWI)

EPI or GRE (hence susceptibility artifact). Molecutes move randomly in fluid (Brownian motion), variably restricted with cell membranes, vascular structures, density of cells, axons. First gradient dephases spins, second rephases (if no relative movement); free movement in eg CSF -> loss of coherence and signal loss. DWI signal S(b) = S0 x e -bD, where S0 is intensity at b=0, D = diffusion coefficient. At least 3 sets of orthoganal acquisitions performed (AP, SI, RL) and averaged (insensitive to anisotropy). Parameters include direction of diffusion gradients (1 or all three x/y/z planes) and b value (sec/mm2, measure of diffusion gradient strength, duraltio nof pulse and time of diffusion measurement; higher b values the more diffusion weighted). Difussion restriction in oedema/swelling ( failed sodium-potassium pump eg acute infarct, nonhaemorrhagic trauma), abscess/pus, epidermoid cyst, hypercellular tumour (eg medulloblastoma), MS. Increased diffusion (iso-DWI, high ADC) in vasogenic oedema, CSF collections (eg arachnoid cyst). Normal diffusion but high T2 (T2 shine-through eg subacute infarct) high DWI and iso-ADC. b0 images are fast T2WI, useful for uncooperative pts and susceptibility detection.

DWI is intrinsically T2W, hence high signal areas may show T2 shine-through, but ADC usually not reduced. b0 is when diffusion gradient is off (bright CSF, bright fat). ADC map (apparent diffusion coefficient) – one set of images with DWI (b >800), the other without (b0, fast to perform, can be used as cheap gradient); a negative logarithm of the ratio of the two is obtained; areas of restricted diffusion appear dark and used to counter T2 shine-through (high S0).

Diffusion restriction can be due to increased viscosity (slowing down water movement), eg abscess, haematoma; or spatial restriction (water movement is fast but barrier limited) eg cytotoxic oedema, epidermoid (organised layers of keratin). Fast diffusion time (short TE) means DWI will not pick up spatially restricted lesions (eg infarcts will not be bright), but less motion artefact.

Acute stroke = bright DWI (within minutes of irreversible damage, most sensitive test for infarct), dark ADC; subacute stroke bright DWI, bright ADC; old stroke dark DWI, bright ADC. Transient ischaemia (TIAs) do not usually show diffusion restriction.

Tensor diffusion imaging/tractography (TDI, white matter mapping) – In elongated cell processes (eg axons), water can diffuse more freely down the tube than sideways (difusion anisotropy). Diffusion gradients applied in at least six directions (rather than 3 as in standard DWI), identifying location, orientation and directionality of white matter tracts. Tensor = magnitude and direction of diffusion w 3 x 3 matrix; Eigenvalue = mathematical property of tensor/vector. Tractography = postprocessing to create images representing axonal fibre tracts from diffusion tensor data, usually in colour with tracts orientated L/R red, CC blue, AP green. Isotropy = uniformity of molecule, absence of polarity, anisotropy opposite; fractional anisotropy = degree w 0=isotropy, 1=full anisotropy. Increased myelination incresases anisotrophy. Used for stroke, tumour, trauma, demyelinating disorders.

Intravoxel incoherent motion (IVIM) – Bi-exponential model of ADC. On a signal vs b-value graph, with increasing b-values signal will initially drop very rapidly (D*), then later drop slower (D). D* (pseudo-diffusion co-efficient) is mostly due to capillaries, with b <160 s/mm2. True diffusion of the tissues is the slower decaying D. f, the perfusion fraction can then be calculated, represents proportion of tissue occupied by capillaries. Requires acquisition at multiple b-values.

Perfusion Scanning

High-volume gadolinium bolus (3-5mL/sec) with multiple rapid EPI in several locations, where Gad appears dark on the GRE T2*W. Most reliable on 1st pass. Inflow curve plots intensity vs time with calculations of time to peak (>6sec indicates ischaemia) and MTT. Able to identify ischaemic penumbra (brain at risk of infarct, still DWI negative). Increased CBV in a tumour correlates with angiogenesis, hence tumour grade.

Spin tagging (arterial spin labeling ASL) – more sophisticated, doesn’t use Gad, more accurate for absolute CBF and can be used repetitively. Labels inflowing blood w saturation bulse. Low SNR and sussceptibility artifacts.

Functional MRI (fMRI)

Evaulates regional neuronal activity. Uses BOLD (blood oxygenation level-dependent) where deoxyhaemoglobin is paramagnetic, shortening T2 and T2* hence lower signal. Increased metabolic activity -> increased blood flow -> washout of deoxyHb, increased oxyHb:deoxyHb ratio (BOLD effect). Mapped onto anatomical information, with comparison of images taken with patient is resting, sensory stimulation, motor activity, higher cortical task. Used for seizures, planning for brain tumour surgery.

MR Spectroscopy (MRS)

Proton/hydrogen MR spectroscopy – distribution of metabolites based on chemical shift of protons. Single-voxel sepctroscopy uses 1 cm2 sample voxel with each metabolic peak charatarised by resonance freuency, height, width and area. Normal stair-step relationship of choline (Cho, 3.2ppm), creatine (Cr, 3.0ppm) and N-acetyl aspartate (NAA, 2.0ppm). MR spectroscopy imaging (MRSI) creates metabolite maps with multiple voxels in matrices and slices, but requires time and large post-processing requirements.

  • Myoinositol (mI) – marker for Alzheimers disease
  • Choline (Cho, 3.2ppm) – marker of cell membranes synthesis and degradation (hence cellular density), increased in high cellular turnover. High choline-to-creatine ratio in tumour or infection, higher grade tumour, very high in meningioma.
  • Creatine and phosphocreatine (Cr, 3.0ppm) – evenly distributed in may types of cells, assoc w cellular energy metabolism, reference standard. May be increased in hypometabolic states, decreased in hypermetabolic states.
  • Glutamine and glutamate residues (Glx, 2.3ppm)
  • N-acetyl aspartate (NAA, 2.0ppm) – compound specific to neurons, marker of neuronal integrity. Reduced NAA-to-creatine ratio w neuronal death. Focal decrease in mesial temporal sclerosis, infarcts, SOL replacing brain, abscesses, mets (lower than primary brain tumours which infiltrate rather than replace brain), radiation, seizures. Global depletion in MS, alzheimer’s. Markedly elevated NAA in Canavan’s disease (defect in enzyme metabolising NAA).
  • Lipids (1.0-1.5ppm) – often in mets, acute MS plaques.
  • Lactate peak (>1.3ppm) – not normally detectable, doublet peaks in anaerobic metabolism. Acute ischaemia, seizures, infarction, high grade tumours, severe infections.
  • Necrosis peak – nonspecific, in malignant tumours, infection, some active demyelinating lesions.
  • Alanine – high in meningiomas.

Phosphoros MRS requires larger volumes with lower SNR.

Magnetisation Transfer (MT) Imaging

Pools of protons include free mobile unbound (water), immobile restricted motion (invisible to MR), and boundary layer protons where exchange of MT occurs. Contribution of MT from immobile proton pool (cell membranes, macromolecules, lipid bilayers) determined by change in signal between being suppressed (off resonance pulse) and not suppressed.  After saturation pulse with large amplitude away from water and fat frequencies (targetted to macromolecular pool which has broad frequency) the compounds exchange magnetisation with adjacent water. Magnetisation transfer ratio (MTR) = 1 – MTs/MTo (saturation pulse on/off). Hence such things as demyelination (MS) there is loss of myelin, less MT and reduced MTR. Scans with and without suppression analysed to assess contribution of macromolecular pool. MT suppression pulse can also be used to reduce background signal in MRA, white matter (suppresses more than grey matter) for better conspicuity of enhancing white matter lesions.

MR Angiography (MRA)

Cervical MRA tends to overestimate moderate stenosis, esp 2D TOF.

Time-of-Flight Imaging (TOF)

Uses difference between longitudinal magnetisation of nonsaturated inflowing blood with stationary saturated tissue (flow-related enhancement). Multiple RF pulses are applied with short TR to saturate spins in stationary tissue, where inflowing blood is unaffected (flow-related enhancement). May be acquired T1WI 2D (more sensitive to slow-flow) or 3D (better for flast flow, very thin-section MRAs w ability to rotate in space, less degraded by motion). Usually cervical done as 2D with small volume 3D of carotid bifurcation; intracranial done as 3D TOF. May be done as MRA (saturating distal/superior veins) or MRV (saturating proximal/inferior arteries). Any ares of intrisic high T1 (subacute haematoma, fat) incl subacute intramural clot in dissections and venous sinus thromboses will appear bright. Marked improvement with 3T coils due to increased SNR and increased T1 (hence easier to suppress background signal), hence ability to assess small branches and vasculopathies.

MOTSA = multiple overlapping thin-slab acquisitions = thinner slab than 3D so less affected by distal saturation

images reviewed as individual sections or MIPS (maximum intensity projections)

Phase-Contrast Imaging (PC)

Uses change in transvere magnetisation (phase shifts) when flowing protons undergo change in gradient strength. Bipolar flow-encoding gradient pulse applied to tissue between initial excitation and readout pulse, hence phase shift induced in blood, but not stationary tissue. Two acquisitions performed with velocity-encoding pulses (VENC) of opposite polarity, with velocites that angiogram is sensitive to (neuro 30cm/s arterial, 15cm/s venous). These two images are subtracted; one sensitised to flow in one direction and the other in a different direction with stationary tissue signal being subtracted out (more completely than TOF). Also enables velocity information by flow-encoding gradients adjusted to range of velocities interest in (ie just slow-flow venous/CSF or fastlfow arterial), or dierectional encoding (eg R->L is bright). Can be done 3D or 2D. 3D has better SNR due to thick-slab excitation, thinner sections with less saturation hence brighter signal in vessle; but longer TE increasing dephasing, longer scan and less sensitive to slow flow (saturation within a larger volume). Les susceptible to intrinsic T1 (ie venous subacute thrombus) hence used more for MRV.

high signal is flow S/R/A -> I/L/P

Contrast-Enhanced MRA (CEMRA)

Gadolinium injected (~2mL/sec) with sequential MRAs (creating different phases) or after ‘SmartPrep’ or test bolus. Circulation time takes ~10-25sec. Used in conjuction 3D TOF. Shows slow flow better than TOF, shows thight narrowings better, isn’t susceptible to flow artifacts or turbulence, but timing critical to avoid venous contamination. Coronal scanning allows with large FOV allows ideal evaluation of vessel origins. Also need to do pre-contrast T1, as bright on MRA may be fat, haematoma, high protein or cholesterol.

Fat-Related Imaging

Fat Signal Suppression

spoiler/crusher gradient immediately following an RF pulse shifts NMR of fat to zero Mz; affected by magnetic field inhomogeneities, incomplete fat saturation from imperfect selective excitation

Fat saturation – saturation pulse at resonace freuency specific to fat (suppresing macroscopic fat within fat cells), effective w CE, but highly sensitive to magnetic field inhomogeneities and mis-registration artifacts, doesn’t work well with low-field strength.

STIR suppresses fat as well as tissues with very short T1 (including gad, haemorrahge, mucoid tissue, proteinaceous fluid). It is insensitive to inhomogeneities in magnetic filed.

water-excitation so only tissues containing water have Mxy, with fat zero Mxy (eg spectral-spatial RF pulse)

In-and-Out-of-Phase Imaging

used to detect microscopic fat based on different chemical environments causing different ω0 of fat and water. Usually spoiled GRE used with imaging of protons in phase (TE 4.2ms @1.5T) and out of phase (TE 2.1ms @1.5T). Water and fat mollecules are in-phase at even multiples of 2.3ms (at 1.5T); out-of-phase at odd multiples. Opposed phase/out-of phase (OP) detects small amounts of intracellular fat, subtracting out water signal; prominent ‘indian-ink’ artefact. In-phase (IP) images add signal from fat and water. Adipose tissue (macroscopic fat) is minimaly changed between OP and IP. Tissues with low fat contenet but high water content (microscopic fat) show prominent loss of signal on OP compared to IP images (adenomas, hepatic steatosis).

MRI Reconstruction

Imaging Techniques

sequential point (each voxel acquired independently), line, planar =2-dimentional imaging or volume = 3-dimensional imaging

planar and volume imaging now used with projection reconstruction (using back-projection, prone to motion artefacts) or 2D/3D Fourier transform (newer machines)

2D Fourier Transform (FT) = Phase Encoding Method

linear field gradients on the main field, causing ω(x) = ω0 + γGxx; Gx = magnitude of gradient field

slice selection gradient excites protons in a thin slice with Gz and appropriate RF frequency bandwidth; increasing Gz decreases the Δz (slice thickness), increasing Δω (bandwidth of RF pulse) increases thickness

Projection Slice Phase Freq
Axial (body) Gz Gy Gx
Axial (head) Gz Gx Gy
Coronal Gy Gx Gz
Sagittal Gx Gy Gz
Gy in the smallest direction to reduce imaging time

phase encoding/evolution period with Gy applied to increase ω for a short time and cause protons to be out of phase along the y-axis

readout = frequency encoding gradient, Gx applied during FID being recorded hence value of ω dependent on position along the x-axis

a large number of FIDs with different magnitudes of phase encoding gradients is collected which undergo FT to produce a k-space matrix; the k-space undergoes a second FT to extract phase encoding information before inverse FT to produce the image matrix

Gx and Gz remain at a constant amplitude, whereas the amplitude of Gy varies; for a 512 x 512 matrix Gx needs to be sampled 512 times, and 512 values of Gy performed to obtain separate FIDs

multisection = interleaved multislice imaging = changing Gz to excite a different slice while waiting for T1 relaxation; number of slices imaged simultaneously depends on TR and time taken to collect information following each excitation; increases movement artefact

3D Fourier Transform

a weak gradient is applied to select a slab of tissue and phase encoding is applied in two directions (Gy, Gz) simultaneously or sequentially; Ny x Nz number of different phase combinations and FIDs are required for 3D-FT

disadvantages: interleaved multislice imaging is not possible, and this technique can take a long time (can be reduced by decreasing TR [thus less contrast] or lower flip angle)

advantages: improved SNR (less noise due to large volume of tissue), less interslice cross talk, very thin contiguous slices possible (minimal partial volume effects); images can be generated in any plane without loss of resolution, can be used for surface/volume rendered 3D images

Image Quality

if receiver coil is dominant source of noise, SNR Ho7/4 (small biological samples), but SNR limited by thermal noise from tissue RF attenuation, hence SNR Ho (large biological samples), except in high H0 where high G fields result in less than expected SNR

SNR intrinsic signal intensity x voxel volume (slice thickness x pixel size2) x √N signal averages

procedure time is higher with slice thickness signal averaging (identical combinations in a slice to increase SNR)

tissues with short T1 and long T2 have high signal thus high SNR

shorter TR (means only partial recovery of Mz) and long TE (smaller Mxy) hence lower SNR, however high contrast betw T1 and T2 respectively

the higher the receiver bandwidth (with increased gradient), the lower signal at a certain frequency hence lower SNR; , but increased spatial resolution

surface coils for direct placement on the body improves spatial resolution due to smaller FOV (with same matrix size) and higher SNR (due to noise being only from the imaging volume)


errors in spatial encoding, worse in the phase encoding direction

distortions in gradient fields and variations in magnetic susceptibility causes spatial distortions

distortions in Hloc from metal implants

non-uniformities in H1 most obvious with surface coils at the periphery of FOV

bounce point artefact – in IR when TI = 0.69 x T1 of that tissue causing a null signal

zipper artefact – dashed lines, mostly caused by inhomogeneities in magnetic field from interference RF field

susceptibility artefact causes distortion in local magnetic field, thus change in ω0 and drop out in signal

motion artefacts

worse in phase encoding direction due to change in location betw cycles causing a streak/ghost in final image

cardiac gating (also reduces CSF pulsation in C/T-spine)

regional resaturation with 90° pulse selectively applied to moving tissue to prevent of from producing a signal

chemical shift

H0 causes induces a current in the electron cloud  local magnetic field (Hloc) opposed H0

Heff = H0 – Hloc = H0 (1 – σ); σ = shielding constant

electron cloud is dictated by the nature of chemical bonding

since ω0 is determined by Heff, chemical shift is quoted as fractional shift in ω0, ie σ1 – σ2 in ppm (parts per million); shift betw fat and water ≈ 3.3 ppm but increases with H0

‘misregistration’ of ω0 thus x/y/z gradient does/doesn’t affect the protons

causes apparent displacement of fat/water boundary; causes dark and bright rims at opposite edges of eg kidney in frequency encoding direction

controlled by increasing receiver bandwidth (but also reducing SNR)

aliasing = wrap-around artefact

portions of the object extend outside the reconstructed FOV and the FID is not sampled frequently enough

worse in the phase encoding direction

overcome by increasing size of acquisition FOV (with pixel size maintained) and irrelevant part of image discarded after processing; or presaturation applied to area outside FOV

effects of flow

paradoxical enhancement = bright blood imaging

flow void = black blood imaging

in IR if TI <T1 of blood, a proton moving into the area after the 180° pulse will appear bright due to the single 90° pulse

in SE if TR <T1 static blood is dull, slow blood or short TE is bright (fresh blood strong Mz as not exposed to previous 90°), fast or long TE is dark (not exposed to rephrasing 180° hence weak signal); if TR > T1 static/slow blood is bright, fast blood is increasingly dark (doesn’t undergo 180° rephasing)

signal from blood may be incorrectly positioned in phase encoding direction, reduced with ECG gating and fast imaging methods using flow suppression

MRI Instrumentation


properties: field strength, homogeneity, temporal stability, bore size (1m, reduced to 60cm due to gradient and RF coils)

field strength: fast imaging 0.5-1.5T, spectroscopy using chemical shift >2.0T

field inhomogeneity > 10ppm results in severe distortion and blur, now better than a few ppm; optimal homogeneity achieved automatically with shimming

categories of magnets: resistive (low strength), permanent (low strength), superconducting (up to 10T, liquid He surrounding windings of niobium titanium which is superconducting at 4K/-269°or results in quench = catastrophic thermal/mechanical problems)

low strength fields (0.2-0.3T) have low SNR but cheaper, safer, not as claustrophobic, allow kinematic imaging of joints and allow possibility of interventional/surgical imaging

shim coils are active or passive, adjusting main magnetic field to improve homogeneity, inside main magnet

Gradient System

set of three orthogonal DC coils inside main magnet, typically 10-30mT/m, rise time 0.1-0.6ms

duty cycle = percentage of time the gradients are switched on during acquisition cycle

slew rate = ratio of gradient field strength to rise time (for EPI ≈ 300T/m/s)

RF Coils

two saddle-shaped coils inside main magnet coils with field perpendicular to H0; RF homogeneity important for image quality

need to be tuned prior to each acquisition

the transmitter supplies power for the pulses and is applied for ≈0.23ms for 90° pulse

separate surface receiver ± transmitter coils may be employed eg birdcage (head), single-turn solenoid (extremities and breast), saddle coils

the receiver amplifies the microvolts into useful levels


FIDs pass through an ADC and amplitude sampled at 128-1024 intervals

the computer controls all aspects of the imaging process, performs FTs etc

MRI Safety

pregnancy: may be used if other non-ionising forms are inadequate; no observed deleterious effects but some centres exclude in 1st trimester

Static Magnetic Field

no permanent observed effects of 2T fields

contraindications include implants with electrical circuitry, pt requiring life support equipment, ferrous clips or implants, except for low strength extremity scanning

safe for stents in ≤ 1.5T

Time-Varying Magnetic Field

transient effects: elevation in T-wave of ECG signal, nerve stimulation including pain (from switching of gradients)

noise caused by switching of gradients


RF pulses should not constitute hazard (are 2 orders of magnitude < diathermy)

SAR = specific absorption rate = rate that energy from RF pulse is dissipated in tissue per unit of mass; levels set so body temperature increases do not exceed 1°C; α2.H02.D; α = flip angle, D = duty cycle (inverse TR)

electrode contact with skin may act as antenna for RF signals and cause burns


shielding from fringe/stray magnetic field outside bore are passive (slabs of ferromagnetic material eg iron; steel/copper in walls; also reduces impact of external influences) or active (additional coils built into magnet assembly)

Faraday cage = copper mesh in walls/windows that stop RF radiation from outside

Principles of MR Interpretation

Mineral-rich structures (bone, calculi) and colalgenous tissue (ligaments, tendons, fibrocartilage, fibrosis) are low in water content and lack mobile protons, hence low signal on all sequences. Bound water (restricted motion) is mostly in intracellular fluid, owing to hydrogen bonding w proteins. Free water (unrestricted motion) in extracellular fluid (and some intracellular) is low T1 and high T2.

Tissues with large amounts of free water include CSF, cysts, bladder, bile, kidneys (urine), ovaries & thyroid (fluid-filled follicles), spleen & penis (stagnat blood), prostate testes & seminal vesicles (fluid in tubues), oedema. Most neoplastic tissues have increased ECF and proportion of free intracellular water.

Proteinaceous fluids have high/shorter T1 and shorter T2, but T1 effect dominates even on T2 so intermediate T1, high T2; tissues include synovial fluid, complicated cysts, abscesses, nucleus pulposis, pathologic fluid colelctions, necrotic areas within tumours.

Soft tissues have predominantly intracellular bound water (shorter T1 and T2) with low T1 and low/intermediate T2, including liver, pancreas, adrenals, muscle. Intracellular protein synthesis shorteens T1 even more, hence muscle (lses active protein synthesis) is lower T1 than other organs. Hyaline cartilage has predominantly extracellular water, but is extensively bound to mucopolysaccharide matrix hence intermediate signal.

Fat has short/bright T1 and shorter T2 than water. On strongly T2WI fat is intermediate/low. On images with lesser T2 weighting T1 effect predominates, hence fat is intermediate to high signal.

Effects of Flow

Paradoxical enhancement = bright blood imaging, flow void = black blood imaging. Slow moving blood (eg spleen, venous plexuses, cavernous haemangiomas) is low T1/high T2 from extracellular free water.

In IR if TI <T1 of blood, a proton moving into the area after the 180° pulse will appear bright due to the single 90° pulse.

In SE if TR <T1 static blood is dull, slow blood or short TE is bright (fresh blood strong Mz as not exposed to previous 90°), fast or long TE is dark (not exposed to rephrasing 180° hence weak signal); if TR > T1 static/slow blood is bright, fast blood is increasingly dark (doesn\’t undergo 180° rephasing)

Signal from blood may be incorrectly positioned in phase encoding direction, reduced with ECG gating and fast imaging methods using flow suppression


See Cerebrovascular Disease