Fundamentals of MR Imaging

Transcription

1 Fundamentals of MR Imaging Shantanu Sinha. Department of Radiology UCSD School of Medicine, San Diego, CA

2 Background References: R.B.Lufkin, The MRI Manual (2nd Edition). Web: Donald G. Mitchell MRI Principles Stark and Bradley. A tiny little bit of history! MRI as a Radiologic Imaging Modality.

3 Basics of Nuclear Magnetic Resonance 1 Magnetic Moment, m Constituents of Matter: Atoms Electrons + Nucleus Angular Moment, I Absence of Field Nucleons: Protons + Neutrons e.g. 1 H (simplest) If no. of nucleons unpaired: Angular Momentum, I, i.e. wobbles about axis. If combined with +ve charge, Magnetic moment, m. Gyromagnetic Ratio, m = g * I gˆ

4 Basics of Nuclear Magnetic Resonance 2 Human body is made up of billions of these tiny magnets (and others as well) Randomly oriented in the absence of any field. Nuclei Angular Momentum, I Isotopic Abndnce Freq. At 10 KG g Gyro-M Ratio Relative Sensitivity for equal # of Nuclei Overall 1 H ½ C ½ N P ½ I for: 23 Na =3/2; 25 Mg = 5/2; 43 Ca = 7/2 There can be MR of the electron as well, EPR

5 Basics of Nuclear Magnetic Resonance 4 Classical Mechanics: Energy Absorbed and Angle of Orientation, can vary continuously. E = m.b 0 = m.b 0.cos(a) Quantum Mechanics: Discrete Zeeman Splitting DE = ghb 0 = Larmor Precessional Frequency: Photon Energy E p = hw 0 w 0 = gb 0

6 The Intrusion of Radio Frequency 9 M z + Tipping the Spins: a 1 = wt 1 = gb 1 t 1 p/2 pulse a = 90 o p pulse a = 180 o t 1 B 1 RF pulse B 1

7 How the signal is generated B0 B0 N transmitter excitation pulse S Signal Phase Voltage Free Induction Decay (FID) Signal frequency Signal Amplitude time

8 Voltage Detection of Signal After Tipping of Spin 10 B 0 Spin System 2. A 90o RF pulse tips Mz to X-Y plane. Mz now rotates in X-Y, transverse plane, perpendicular to plane of detector coil. RF p/2 Detector Coil 1. Most spins are aligned along Z, forming M z. Since M z is in plane of detector coil, no Voltage is generated. 3. As it rotates in transverse plane, M x produces an alternate, sinu-soidal current in coil in the X-Z plane. Time

9 B0 B0 N transmitter excitation pulse 9

10 M z (t) T 1 and T 2 Relaxation o Flip by RF Signal Detection Dephasing B 0 Z M x (t) ~ e t/t2 M z (t) ~ M 0 (t){1- e t/t1 } M z M 0 X Y Time

11 B0 B0 B0 time T2 relaxation process Magnitude of the magnetization component in the plane perpendicular to B0 decays to zero Voltage Free Induction Decay (FID) Signal frequency Signal Phase Signal Amplitude 11

12 B0 B0 B0 N S N S N S time T1 relaxation process Magnitude of the magnetization component parallel to B0 grows and ultimately reaches an equilibrium 12

13 Excitation, Evolution and Relaxation 12 M z a 0 RF Pulse Y a 0 M X Y X T1 Relaxation X M Y X Y T2 Dephasing

14 Spin Echo w+dw w+dw w-dw w-dw RF p/2 (90 o ) p (180 o ) Spin Echo Signal FID TE/2 TE/2 Time TE

15 T 2, T 2 *, Multiple Echoes, Measurement of T 2 T 2 *: Due to inherent magnetic field inhomogeneities, dephasing occurs much more rapidly than T2, much more rapid decay of T2 curve. Multiple Echoes: 1 T 2 * 1 = + Can create several echoes, by using repeated 180 o RF pulses to flip dephasing spins to the other side, and allowing them to rephase. T 2 1 DH This is multiple echoes, (Double Echo, when only two). RF 90 o 180 o 180 o 180 o TE1 ~20ms M z (t) ~ M 0 (t)*e t/t2 TE2 ~80ms TE3 ~120ms Time Measurement of T 2 : Fitting the peaks to the equation will yield T2. Individual Echoes decay at T 2 *, much more rapidly.

16 2D FT and Image Formation: TR Ech Ech o o 180 o 90 o 180 o Ech o Each such X-axis raw spectra, a projection of the Y-column. 90 o 180 o TE Has to be repeated 128~256 times Once entire column for each X-cord collected, a 2 nd FT along Y-axis yields Y- axis information. Total Scan Time = N pe * TR*N avg

17 Contrast Mechanisms (T1, T2 and r) T1, Spin-Lattice Relaxation: Time taken for excited spins to relax back to ground state. White matter white, gray matter gray, CSF black. T2, Spin-Spin Relaxation: Time taken for excited spins to relax from ordered to disordered state. White matter black, gray matter gray, CSF white. r, Proton Density: # of (fluidic) protons/unit volume Contrast intermediate, with solids such as bone appearing black. 17

18 Image Contrast - Proton Density and T2 I z (t) ~ r e t/t2 {1- e t/t1 } TR/TE: 2000/25 TR/TE: 2000/50 TR/TE: 2000/75 TR/TE: 2000/100 TR/TE: 2000/150 TR/TE: 2000/200 18

19 Generation of T1-wtd images T1-weighted Signal Intensity WM Contrast-enhanced TR/TE: 120/20 TR/TE: 240/20 GM CSF TR/TE: 500/20 TR/TE: 1000/ TR (msec) TR/TE: 2000/20 TR/TE: 4000/20

20 Effect of TE/TR on visualizing pathology TR/TE: 600/20ms TR/TE: 2000/30ms TR/TE: 2000/90ms

21 T1, T2 of Different Tissues

22 Inversion Recovery Measurement of T 1 Z +M 0 Z +M 0 Fat Muscle Y Fluid, CSF Time X -M 0 X -M 0 M z (t) ~ M 0 (t){1- e t/t1 } RF Echo 180 o TI, Inversion Time 90 o TE/2 180 o TE/2 STIR, Fat suppression FLAIR Inversion Recovery: First tip by 180 o, then let decay (longitudinally) for TI (~300ms) Then add on a standard spin echo (90 o o ). Can repeat for several TI s. Curve of spin echo heights will give T1 values. Images are heaviy T1-weighted.

23 Enhanced Contrast: Gadolinium-DTPA: T1 contrast enhancement by IV injection of exogenous contrast agent. (Image: 14 yr. Old HIV patient). lesion Inversion Recovery (IR): Maximization of T1 contrast by special pulse sequence. (Image: Normal Brain). Fluid Attenuated IR (FLAIR): Suppression of bright CSF in T2- wtd. image to better visualize periventricular lesions. (Image: 14 yr. Old HIV patient). lesion

24 Magnetic Field Gradients : If the two vials are in same magnetic field, B 0, they both resonate at: w 0 = g*b 0 since they are in the same magnetic field They produce FID s at the same frequency. If the magnetic field is made to vary linearly along the X-axis as: B(x) = B 0 + (db/dx)* x, and, in frequency, w(x) = g*{b 0 +(db/dx)*x} = w 0 + g*(db/dx)*x, Then frequency of spin becomes dependent on it s X-spatial coordinate. Signal becomes spatially encoded, and one has mapping! The spins in the two vials produce FID s at two different frequencies since they are in different magnetic fields Similar gradients can be switched on each of the 3 physical axis's, to encode in all three directions.

25 Fourier Transform: Domain Domain Data is collected conveniently in one domain (time). Then Fourier Transformed to a space where it is easier to analyze (Frequency) The two data are connected to (and can be obtained from) each other by the Fourier Integral. Examples: Sound Perception, Music uses FT Visual Perception: Cannot FT. Useful Concepts: Dt Dwell Time T N*Dt Acquisition Time.

26 Magnetic Gradient Fields G X B Z X G X B Z Y G X B Z Z w=gb RF/Echo Slice Frequency Phase

27 Slice Selection Gradients with RF on: B 0 Left Right Slice Select Gradient p/2 (90 o ) 180 o 90 o Slice Selection: Spins aligned along B 0 Superimpose Slice Select gradient, db/dx: Linear variation of mag nc field w/ one axis Switch on RF 90 0 pulse Spins only within a slice are flipped by 90 0 to the X-Y plane SSG also switched on during pulse, so that same spins are flipped over. RF Pulses Slice Select Gradient

28 Slice Selection: High Position of Slice: Change frequency of RF f i = g*(b 0 + (db/dx)*x i } High Thickness of Slice: Change spread or bandwidth of RF pulse Or, can change (db/dx) Thin Slice Low Signal Frequency Frequency Low Low B 0 in presence of S/I magnetic field gradient MR signal frequency MRI of the selected slice

29 Frequency Frequency A few quick questions on Slice Selection: Position of Slice: Thickness of Slice: RF Pulse F1 RF Pulse Higher Gradient Lower Gradient RF Pulse F2 Slice Position (x) Thicker Slice Slice Position (x) Thicker Slice 1) At 1.5T, if Slice Thickness 10 to 5 mm, Gradient Strength? If Magnetic Field 1.5T 3T, for same change in Sl.Thk, Grad Strength? If Magnetic Field 1.5T 3T, how much will frequency need to be changed if going from left ear to right? For (i), what if changing from H nuclei to P?

30 B 0 Phase Encoding Gradient: Anterior q 1 q 2 q 3 q 1 q 2 Phase Encoding: For the spins selected by the SS The phase encoding gradient, db/dz, switched on for a short time Spins in different layers along the Z axis, rotated at different speeds, for the SAME time, Posterior q o But to different angles, q 1,q 2, q 3 These phase differences, or encoding, by which the spins are tapped, are remembered till the signal is finally detected. Is finally used in the Fourier Transform 90 o RF Pulses Phase Encoding Gradient

31 Read-Out Gradient: B 0 Superior f 1 f 2 f 3 f 4 f 1 f 3 Last gradient, db/dy, to be switched on, along the final, Y-axis. During detection of signal. Signal from each vertical slab, is at a different frequency, f 1, f 3, f 5. Inferior f o f 5 since they are in different magnetic field Final signal is a convulation of all these different frequencies, with phase differences from PE, and from protons selected by SS. 90 o RF Pulses Read-Out Gradient Echo

32 2D FT and Image Formation: TR Ech Ech o o 180 o 90 o 180 o Ech o Each such X-axis raw spectra, a projection of the Y-column. 90 o 180 o TE Has to be repeated 128~256 times Once entire column for each X-cord collected, a 2 nd FT along Y-axis yields Y- axis information. Total Scan Time = N pe * TR*N avg

33 K-Space Reciprocality with Physical Image Space: (A) (B) (C) FT of k-space data yield MR images. Different parts of k-space influence appearance of MR image in different ways. (A) FT of entire K-space Image w/ good contrast and resolution. (B) Only center of K-space Contrast of image (C) Outer lines Spatial resolution information, edges and contours of organs and other minute structures.

34 FT of K-Space to Physical Space 2 nd Example: K-Space Data (A)All points included (B)Only central lines Contrast (longwavelength) (C)Outer Lines Edges

35 Reducing Scan Time Lower # of PE levels: Reduce # of PE levels: Reduce the total range of PE levels, Symmetric about zero Increment between levels remain the same. Zero-fill to original matrix size. Reduces total scan time: By 256xTR to 192xTR to 128xTR, FOV remains the same since, zero filled to original matrix size Resolution along PE direction, reduces in the ratio of 1:0.75::0.5 Resl n along RO remains same./ Downside: Possible Wrap-Around

36 Effect of Image Matrix Steps on SNR & Image Quality: 32x256 matrix 64x256 matrix 128x256 matrix FAST Spin Echo # of PE most important Improves in-plane spatial resolution in PE direction Increases total imaging time Decreases voxel volume Decreases SNR per voxel Decreases extent of truncation (Gibb s) artifact 192x256 matrix 256x256 matrix 512x256 matrix 256x32 matrix 256x64 matrix 256x128 matrix

37 Wrap Around Artifact -- Phase Direction: Number of Phase Encoding Steps: Each PE step takes TR amount of time. Larger the number of PE steps, better the resolution in that direction But, increases the total scan time. WRAP-AROUND ARTIFACT: If sufficient number of steps not acquired to cover range of anatomy along that direction Can be eliminated by increasing # of PE steps, Can change PE/RO directions to minimize effects. No Wrap Around Wrap Around Artifact

38 Motion Artifact Phase Direction: Motion Artifact: Since each PE line takes TR amount of time, Anatomy can be in different positions Physiologic Motions Motion artifacts registered along PE line. Can be eliminated by: Short scan time, Switch axis, Gating: Cardiac, Respiratory

39 Effect of Phase Encoding Steps on SNR and Image Quality: T1W: SE 500/26 ms N pe = 32 N pe = 64 N pe = 128 # of PE most important Improves in-plane spatial resolution in PE direction Increases total imaging time Decreases voxel volume Decreases SNR per voxel Decreases extent of truncation (Gibb s) artifact N pe = 192 N pe = 256 N pe = 512

40 15 cm FOV Effect of Changing FOV, w/ matrix constant: 60 cm FOV 50 cm FOV 40 cm FOV 30 cm FOV 25 cm FOV 20 cm FOV 10 cm FOV Matrix size is kept constant at 256x256: As FOV is decreased: Spatial resolution increases due to smaller voxel dimensions in both inplane directions, SNR per voxel decreases due to smaller voxels Aliasing (wraparound) artifacts begins to occur when signal producing material is outside FOV in the PE direction Can be eliminated by over-sampling SNR increases as square of FOV

41 Effect of Bandwidth on SNR & Image Quality: BW = 2.2 KHz BW = 7.8 KHz BW = 2.2 KHz BW = 3.9 KHz BW = 15.6 KHz BW = 3.9 KHz Bandwidth (BW) Range of frequencies that the receiver system accepts. As BW is increased, range of frequencies over which signal is received is increased. While signal remains constant, noise increases as square-root of BW. As BW is increased, total duration of signal sampling and dwell time are decreased inversely. To keep FOV in RO direction unchanged, as BW in increased, RO gradient amplitude has to be increased. Chemical shift artifact decreases w/ BW increase. Fig. E and F, show inhomogeneties at low BW, (E), which disappear at higher BW (F). In EPI, BW is increase by order of magnitude.

42 Effect of Slice Thickness on SNR and Image Quality: Slice Thk: 3 mm Slice Thk: 5 mm Slice Thk: 7 mm Signal is proportional to # of protons/ voxel As Sl.thk. Increases, # of protons increases linearly w/ Sl.Thk Partial voluming also increases w/ Sl.Thk. Thinner slices always preferred: since better resolution, Limited by SNR, and gradients. Slice Thk: 10 mm Slice Thk: 15 mm Slice Thk: 20 mm

43 Imaging and Pulse Sequences: D 53: E. Spin Echo Imaging: Echoes generated by a 180o RF. D 54: D. Gradient Echo Imaging: Typical flip angles of 10~40o. D 55: A. 2-D FT has PE. in one direction and RO in other. D 56: B. 3-D FT has PE in both original PE and SS directions, and RO in the 3 rd. D53: C. (1.0T/1.5T)*64= 40 MHz. D54: D. Decreasing slice thickness decreases # of protons creating signal, hence decreases SNR.

44 Contrast: D51: C. Long T1 tends to saturate and short T2 decays the signal too rapidly. Higher signals are produced from short T1 and long T2. D 52: D. T1 ~80 to 3000ms; T2 ~ 10 to 80ms. T2* is reduced from T2. D55: E. Atomic number doesn t play a part. D 56: A. Pure water has a long T1 and long T2, same as CSF. Both can be shortened by D57: C. Gradient fields are used to localize MR signal. D 58: B. RF pulses are used to flip M thru desired angles. D59: A. Shim coils are used to change magnetic fields locally and increase inhomogeneity of field.

45 Artifacts: D57: A. Chemical Shift from differences in resonant frequency of different types of protons.; Zipper Artifact: Typically from motion or Electronic noise; Wrap-around from less # of PE levels. Ring artifact from Gibb s ringing.

46 K-Space Traversal: Standard Spin Echo Standard Spin Echo FAST Spin Echo HASTE Standard Spin Echo Sequence: One line of K-space per TR collected. Sequence repeated to cover K-space. EPI acquires multiple segments of image data from a single SE or GE. Oscillating gradient fast acquisition techniques. Snapshot EPI Acquisition that collects all PE levels from a single echo train.

47 A little bit of Physics (Spin Tags, Spirals etc ) A B C D Ref: 1) VE-PC: Drace (1994); Sinha (2004); 2) Spin Tag: Axel (1989); Ryf (2004); 3) DENSE: Aletras (1999); 4) Single shot VE-PC Spiral, Asakawa (2003). 5) Fig. Bernstein s A. Rectilinear: 1 PE line per R-R interval 128~80 contractions Most dense sampling; higher resolution. B. Spiral Coverage less dense, corners not sampled E F C. Single Spiral: Can cover entire K-space in 1 R-R. Very rapid but low res. D. Interleaved Spirals: N- arms, denser sampling, higher res, but N R-R intervals. E: SPAMM and DANTE RF pulse for producing spin tags F. Spiral Gradient Waveform Combined by us with Spiral Acquisition.

48 Partial Fourier Imaging: Half-Fourier Sampling 3/4-Fourier Sampling Full-Fourier Sampling Partial Fourier Imaging: Collects only part of FID, Does not reduce Image Matrix or Field of View Can be either in PE or RO direction In PE directions, reduces Total Scan Time drastically Generate remaining data by symmetry Save significant time, but Increases artifact In RO direction, can reduce TE

49 Enhanced Contrast: Gadolinium-DTPA: T1 contrast enhancement by IV injection of exogenous contrast agent. (Image: 14 yr. Old HIV patient). lesion Inversion Recovery (IR): Maximization of T1 contrast by special pulse sequence. (Image: Normal Brain). Fluid Attenuated IR (FLAIR): Suppression of bright CSF in T2- wtd. image to better visualize periventricular lesions. (Image: 14 yr. Old HIV patient). lesion 49

50 Contrast from Flow MR Angiography: Utilizes inherent differences in contrast between flowing (blood) spins and stationary tissue. Time-of-Flight MRA: For Visualization of Vasculature. Phase Contrast MRA: Quantifies flow/velocity. (a) Magnitude image, (b) corresponding phase image (note that flow in opposite directions has different signs as in the ascending and descending aorta), (C) velocity profile across aorta reveals laminar flow. 50

51 Phase Accumulated by Moving Spins Distance moved in time dt Phase of Spin Phase acquired by Spin proportional to: 1) Velocity, 2) Gradient (VENC) F 2 F 1 -F 1 Phase difference of moving spins prop. to bipolar gradient strength, duration & separation: These factors define flow sensitivity (venc ) venc should exceed max. velocity in image: o If not, white pixels wrap around as black pixels and vice-versa o If venc too high, vessels not visualized well, overall SNR decreases o Imp. to optimize venc (value close to maximum anticipated vessel velocity) 51

52 2D Time-of-Flight- Principle Based on in-flow enhancement Max enhancement vessel perp. to imaging plane SI of in-flowing blood high, SI of stationary tissue saturated (TR < T1 of tissue) 52

53 Raw Data Set 6 directly acquired images (out of 256) are shown at the level of the carotids 53

54 Post-Processing: MIP Stack of axial slices typically from 2D-TOF or stack of slabs from 3D volume subjected to ray-tracing technique: maximum intensity projections filters strong vascular signals from volume and projects to a single plane brightest voxel is selected along a given direction multiple projection images of the vessels along the slice axis 54

55 & After MIP MIP images in different projections: axial (a), sagittal (b) and coronal (c) 55

56 Phase Contrast MRA RF G z G x G v One image without additional velocity encoding One image with additional velocity encoding Display phase difference between images Phase difference is directly proportional to velocity Phase difference subtracts out off-resonance and other phase effects 56

57 Phase Contrast MRA Phase is proportional to velocity Quantitate velocity from phase images and/or: Construct angiograms by MIP of velocity maps MIP 57