Ultrasound instrumentation and image formation. Lecturer: Chelsea Munding September 28 th, 2017

Transcription

1 Ultrasound instrumentation and image formation Lecturer: Chelsea Munding September 28 th, 2017

2 Outline 2 Review: Ultrasound physics Image formation Transmit block Receive block User-controlled image quality Sources of image degradation Imaging speed Quantifying image quality

3 Review 3 Ultrasound waves may be reflected at the boundary between two different materials (if the acoustic impedance is different) Acoustic Impedance: Z = ρ*c (MRayl) ρ = density (kg/m 3 ) c = speed of sound (m/s) Water (Z=1.54) 87% reflected Steel (Z=45.7) 13% transmitted

4 Review 4 Ultrasound waves may be refracted if they travel into a material with a different speed of sound Edge artifacts due to refraction Crosby and Kendall. (2006) Emergency Ultrasound. Philadelphia: LWW.

5 Refraction and Snell s law 5 Snell s law lets us calculate refraction: Water (c=1540 m/s) θ 1 =20 o Fat (c=1450 m/s) θ 2 =18.8 o

6 Review 6 Small particles (<<λ) cause Rayleigh scattering, producing incoherent interference patterns known as non-specular reflection RCEMLearning (2015). Ultrasound: Physics and basic equipment settings.

7 Review 7 In tissue, this produces a speckle pattern with randomly distributed peaks and troughs RCEMLearning (2015). Ultrasound: Physics and basic equipment settings.

8 Speckle Origins 8 The speckle pattern originates from having multiple scatterers (cells, sub-cellular components) within a resolution cell (e.g. pixel) Each scatterer returns the acoustic signal with variable amplitude and timing (based on size and location) One resolution cell with many scatterers

9 pressure pressure pressure Speckle Origins 9 Superimposing/adding multiple sinusoids with random amplitude and time delay can produce very high or very low values Scatterer = Scatterer 2

10 pressure pressure pressure Speckle Origins 10 Superimposing/adding multiple sinusoids with random amplitude and time delay can produce very high or very low values Scatterer = Scatterer 2

11 Speckle Origins 11 The speckle pattern originates from having multiple scatterers (cells, sub-cellular components) within a resolution cell (e.g. pixel) Each scatterer returns the acoustic signal with variable amplitude and timing (based on size and location) These are summed at the transducer probability of high or low signals depending on spatial arrangement

12 Speckle Origins 12 This is a probability distribution, but a spatial one, not a temporal one speckle pattern consistent when returning to same plane Appearance of speckle directly related to resolution of system and size/distribution of scatterers

13 Review 13 Both scattering and absorption (acoustic energy becoming heat) may contribute to attenuation, causing ultrasound amplitude to decrease as it travels through tissue Hindi et al., RMI 2013:6 p.29-48

14 Attenuation 14 Attenuation depends on ultrasound frequency It is usually reported in units of db/cm/mhz Note: Attenuation is often assumed to be linearly dependent on frequency. Medium Approx. attenuation (db/cm/mhz) Water Blood 0.18 Fat 0.63 Liver 0.94 Along muscle fibres 1.3 Across muscle fibres 3.3 Bone u/bioen508/lecture6-us.pdf

15 Attenuation Example 1 15 Example: How far can a 5 MHz ultrasound wave travel before it attenuates by 20 db (1/100 th the original intensity) in water? Blood? Bone? Water attenuation = db/(cm*mhz) At 5 MHz, attenuation is *5 = db/cm The wave would need to travel 20 db/(0.011 db/cm) = 1,818 cm = m!

16 Attenuation Example 1 16 Example: How far can a 5 MHz ultrasound wave travel before it attenuates by 20 db (1/100 th the original intensity) in water? Blood? Bone? Water attenuation = db/(cm*mhz) At 5 MHz, attenuation is *5 = db/cm The wave would need to travel 20 db/(0.011 db/cm) = 1,818 cm = m! Blood attenuation = 0.18 db/(cm*mhz) At 5 MHz, attenuation is 0.9 db/cm 20 db loss occurs after 22 cm

17 Attenuation Example 1 17 Example: How far can a 5 MHz ultrasound wave travel before it attenuates by 20 db (1/100 th the original intensity) in water? Blood? Bone? Water attenuation = db/(cm*mhz) At 5 MHz, attenuation is *5 = db/cm The wave would need to travel 20 db/(0.011 db/cm) = 1,818 cm = m! Blood attenuation = 0.18 db/(cm*mhz) At 5 MHz, attenuation is 0.9 db/cm 20 db loss occurs after 22 cm Bone attenuation = 20 db/(cm*mhz) At 5 MHz, attenuation is 100 db/cm 20 db loss occurs after 0.2 cm

18 Review 18 There is a trade-off between depth penetration and resolution as both resolution and attenuation increase with frequency Eye ( MHz) Kidney (4 MHz) 1 cm 1 cm Ultrasound/Renal_US_2013.htm

19 Image Formation 19 How are ultrasound images acquired?

20 Transducer basics 21 The active PZT layer resonates when excited by an electrical or mechanical pulse Similar to a bell or tuning fork resonating when struck The resonance frequency of a plate is determined by its thickness: thickness = λ / 2

21 Element Arrays 22 Multiple elements allow creation of focused waves through geometric arrangements or electronic delays Transducer elements Torp, NTNU.

22 Review 23 Ultrasound transducers vibrate to send a pressure wave into tissue, then listen for echoes

23 Image creation sequence 24 We call this received signal an A-line (for amplitude line) Voltage received time

24 Image creation sequence 25 For a standard B-mode, or brightness mode, image we scan the transducer laterally by electronic focusing or use of sub-apertures (subgroups of elements centred over A-line)

25 Image creation sequence The ultrasound system sends an electrical pulse a short sinusoid at a desired frequency amplified to tens (or hundreds) of volts before reaching the PZT element products/ultrasound/systems/iu V peak-to-peak

26 Image creation sequence The transducer element resonates, emits a pressure wave, and changes the content of the signal in a predictable way (based on transducer design)

27 The transmit block 28 Digital signals, mv Controller Transmit Beamformer transmit waveform focusing delays Digital to Analogue Analogue signals, V Analogue signals 10 s to 100 s of V Pressure waves kpa to MPa Amplifier Switch Transmit or Receive

28 Image creation sequence The sound wave reflects off boundaries and scatterers in the tissue (~Pascals) and return to the transducer face Pressure waves Pa

29 Image creation sequence The transducer converts the pressure wave to a voltage signal (~10 s of millivolts) Voltage received (mv) time

30 The receive block 31 Analogue signals mv Switch Transmit or Receive Pressure waves Pa

31 Image creation sequence The analogue voltage signal is amplified to anticipate attenuation (time-dependent) and then converted to a digital signal Voltage received (mv) time Amplification time 0

32 Image creation sequence The analogue voltage signal is amplified to anticipate attenuation (time-dependent) and then converted to a digital signal Voltage received (mv) time Amplified Voltage time

33 The receive block 34 Analogue to Digital Analogue signals mv Digital signals mv Amplifier TGC Analogue signals mv Switch Transmit or Receive Pressure waves Pa

34 35 Voltage received time How do we get from a digital voltage signal to a B-mode image? Lateral distance Depth

35 Basic image processing Beamforming of the A-lines Alignment and summation of echoes occurring from same event, which have been sampled on all transducer elements Time

36 Basic image processing Beamforming of the A-lines Alignment and summation of echoes occurring from same event, which have been sampled on all transducer elements Time

37 Basic image processing Beamforming of the A-lines Alignment and summation of echoes occurring from same event, which have been sampled on all transducer elements Time

38 Basic image processing Beamforming of the A-lines Alignment and summation of echoes occurring from same event, which have been sampled on all transducer elements t 1 t 2 t 3 t 4 t 5 t 6

39 Basic image processing Beamforming of the A-lines Alignment and summation of echoes occurring from same event, which have been sampled on all transducer elements t 1 t 2 t 3 t 4 t 5 t 6

40 Basic image processing Beamforming of the A-lines Alignment and summation of echoes occurring from same event, which have been sampled on all transducer elements t 1 t 2 t 3 t 4 t 5 t 6

41 Basic image processing Beamforming of the A-lines Alignment and summation of echoes occurring from same event, which have been sampled on all transducer elements t 1 t 2 t 3 t 4 t 5 t 6 Echo is smeared due to temporal mismatch

42 Basic image processing Beamforming of the A-lines Alignment and summation of echoes occurring from same event, which have been sampled on all transducer elements t 1 t 2 t 3 t 4 t 5 t 6 t 1 t 2 t 3 t 4 t 5 t 6 Apply a delay to each A-line

43 Basic image processing Beamforming of the A-lines Alignment and summation of echoes occurring from same event, which have been sampled on all transducer elements Apply a delay to each A-line for each point in the field Improves resolution and SNR

44 The receive block 45 Analogue to Digital Analogue signals mv Digital signals mv Amplifier TGC Receive Beamformer Echo re-alignment and summation Analogue signals mv Switch Transmit or Receive Digital signals mv Pressure waves Pa

45 Basic image processing The signal is filtered to remove noise 8. The signal may be passed to processing blocks (Doppler, Nonlinear/Harmonic) covered later 9. The signal is envelope-detected Envelope Voltage received time

46 Basic image processing 47 Envelope detection example: Beamformed RF Data Envelope Detection From J. Dahl, Ultrasound Beamforming and Image Formation, RSNA 2005 Lecture

47 Basic image processing 48 The enveloped waveform is log-compressed and converted to pixel space for display Compression amplifies low values and minimizes high values to emphasize subtleties dominated by strong signals

48 Basic image processing 49 Envelope detection, compression example: Envelope Detection Compressed Data From J. Dahl, Ultrasound Beamforming and Image Formation, RSNA 2005 Lecture

49 The receive block 50 Controller Output Pixel space Audio Image Prep Filtering Motion/nonlinear processing Envelope detection Log compression Banded TGC Analogue to Digital Analogue signals mv Digital signals mv Amplifier TGC Receive Beamformer Echo re-alignment and summation Analogue signals mv Switch Transmit or Receive Digital signals mv Pressure waves Pa

50 User-controlled image quality 51 How do we improve image quality? Selection of transducer (linear, curvilinear, phased) Selection of frequency Output power 2D gain TGC Compounding Focal zone selection and multi-focus

51 Transducer types 52 Philips L9-3 Linear array rectangular footprint Collects axially-oriented A-lines by electrically scanning sub-apertures Mid-high freq, high res., mid-low penetration Vascular, small parts, breast imaging

52 Transducer types 53 Curvilinear array curved footprint Collects sector scans by use of sub-apertures Low-Mid freq, low resolution, high penetration Abdominal, ob/gyn Philips C5-1

53 Transducer types 54 Phased array small rectangular footprint Primarily to image between ribs Collects sector scans by electronic beam steering Low-mid freq, low-mid res., high penetration Cardiac, pediatric, ob/gyn Philips S4-2 Courtesy Excel Medical Imaging

54 Transducer types 55 Others (specialized): Matrix (3D/4D), transvaginal, transrectal, IVUS, etc Where there is a unique imaging constraint, there is a unique probe design BK Ultrasound I12C4F Laparoscopic transducer

55 Selection of frequency 56 Good imaging requires tradeoff of depth penetration and resolution Must choose based on feature size of interest and depth in tissue/knowledge of tissue attenuation

56 Depth Pen. and Resolution 57 Mid-frequency vs. High Frequency Mid frequency (L9-3) High frequency (L17-5)

57 Depth and Resolution: C Center freq. near 3 MHz Low Attenuation (0.5 db/(cm MHz)) No noticeable change High Attenuation (0.7 db/(cm MHz)) No noticeable change

58 Depth and Resolution: L Center freq. near 6 MHz Low Attenuation (0.5 db/(cm MHz)) Slight drop in SNR with depth High Attenuation (0.7 db/(cm MHz)) Max depth 6.5 cm

59 Depth and Resolution: L Center freq. near 11 MHz Low Attenuation (0.5 db/(cm MHz)) High Attenuation (0.7 db/(cm MHz)) Max depth 4 cm

60 Output Power 61

61 Output Power 62

62 2D Gain and TGC 63 2D Gain Increasing or decreasing the attenuation compensation manually applied to whole image Time-Gain Compensation (TGC) Banded gain to adjust image regions individually Bands determined by axial distance (transit time) TGC sliders on a diagnostic ultrasound machine

63 2D Gain and TGC 64 With no 2D gain or TGC adjustment Low dynamic range, uneven image quality

64 2D Gain and TGC 65 2D gain adjustment roughly evens dynamic range Better dynamic range, uneven image quality

65 2D Gain and TGC 66 TGC allows finer adjustment of image regions Better dynamic range, even image quality TGC curve

66 Compounding 68 Compounding generally refers to image smoothing by sampling the same space in different ways and then averaging This increases SNR by averaging out noise and speckle and keeping signal There are a few different ways to do this: Spatial, angular, and frequency compounding Most clinical scanners use a proprietary combination of these

67 Compounding 69 In the most simple schemes, the space is sampled with multiple angles or by translating a sub-aperture (angular/spatial) Transducer + Transducer + Transducer = Transducer Angle 1 Angle 2 Angle 3

68 Compounding 70 This also alleviates artifacts of refraction and attenuation (RSNA artifacts, Lectures 1 and 2) From GE Healthcare whitepaper: Real-time spatial compound imaging in breast ultrasound: technology and early clinical experience

69 Compounding 71 Compounding reduces speckle and smooths the ultrasound images Compounding Off Compounding On

70 Compounding 72 Clearer borders, but what s lost? Compounding off Compounding on

71 Focal Zone Selection 73 The focus determines where you deposit the most energy (i.e. highest SNR) For ideal images, should place focus in alignment with features of interest

72 Focal Depth: L Center freq. near 6 MHz Shallow focal depth Deep focal depth

73 Focal Zone Selection 75 Most platforms offer multi-focus Stitches together separate images taken with different focal zones into one composite Proportional drop in frame-rate From Kremkau 2011, Sonography principles and instruments. Elsevier Press.

74 Image degradation 76 What are sources of imaging noise that degrade the resulting image? Thermal/electronic noise and electrical cross-talk Quantization error Grating lobes and side lobes Speed of sound error Reverberation and aberration from intervening layers Tissue motion

75 Image degradation 77 Thermal/electronic noise and electrical cross-talk Electronic noise, or thermal noise is a random fluctuation in electronics and amplifiers Electrical cross-talk is the coupling of electricity/pressure to neighbouring transducer elements or wires (consequence of miniaturization) Transducer Voltage applied to activate sub-aperture Small stray voltage applied to neighboring elements

76 Image degradation 78 Electronic noise is increasingly apparent with depth as a result of gain compensation

77 Image degradation 79 Quantization error Converting between analogue and digital signals introduces quantization (rounding) error Time delays on transmit Digitization of received voltage From diracdelta.co.uk Quantization error

78 Image degradation 80 Grating lobes and side lobes Side lobes result from a rectangular transducer Grating lobes a result of spacing of elements of transducer Both create off-axis energy that can degrade image From US Artifacts RSNA article Available on course website

79 Image degradation 81 Speed of sound error Incorrect SOS assumption degrades beamforming by introducing aberration in focusing (transmit + receive) Error in distance estimation From US Artifacts RSNA article, Available on course website

80 Image degradation 82 Reverberation from intervening layers A small lie: multi-path scattering doesn t matter Much current research focuses on adaptive beamforming/decluttering to remove these artifacts From JJ Dahl, Reverberation clutter from subcutaneous tissue layers: Simulation and in vivo demonstrations, UMB 2014

81 Image degradation 83 Motion artifacts Must take A-lines sequentially for whole image If tissue motion exists on this time-scale, blurs image Especially problematic in Doppler (Lecture 4) Solutions: Automated motion compensation Robust co-registration Video from

82 Real-time Imaging Speed 84 What determines the overall frame rate? Imaging depth the sound must completely return before acquiring another A-line Number of A-lines the more lines that make up the image, the longer the total acquisition Multiple foci stitched images taken with different focal zones to improve resolution Compounding Averaging of sequential images Interleaving of imaging modes (Doppler/Bmode)

83 Frame rate calculation 85 The speed of sound in tissue is commonly approximated as 1540 m/s. If you want to design a B-mode imaging sequence that consists of 128 scan lines in order to capture features between 0.5 and 5 cm in depth, what is the maximum imaging frame rate that you can achieve? For each A-line, the sound has to travel at least 10 cm (5 cm to max. depth and 5 cm back) Minimum time for each A-line = 10cm 1m s 1540m / s 100cm Maximum frame rate = frame lines 1line s frames / s

84 Frame rate estimation 86 To minimize multiple reflections (late echoes showing up in the next image), it s best to allow extra wait time for all echoes to return before acquiring the next scan line. Assuming that our transducer is sensitive enough to visualize objects at 10 cm, what is the maximum frame rate? (c = 1540 m/s) For each A-line, the sound has to travel 20 cm (10 cm to max. depth and 10 cm back) 20cm 1m s Minimum time for each A-line = 1540m / s 100cm Maximum frame rate = frame lines 1line s 4 60 frames / s

85 Frame rate estimation 87 You find the resulting images poor due to attenuation and electronic noise. To improve this, you acquire 3 different angles and average (spatial compounding). What is the maximum imaging frame rate you can achieve? (c = 1540 m/s, max. depth = 10 cm) Now, each frame consists of 3x as many imaging lines 128 lines at angle lines at angle lines per frame

86 Frame rate estimation 88 You find the resulting images poor due to attenuation and electronic noise. To improve this, you acquire 3 different angles and average (spatial compounding). What is the maximum imaging frame rate you can achieve? (c = 1540 m/s, max. depth = 10 cm) Now, each frame consists of 3x as many imaging lines New frame rate = 1 frame 384lines 1line 0.2 m /1540m/ s 20 frames/ s

87 Frame rate estimation 89 After spatial compounding, you see that features are wellresolved in the middle of the scan depth, but not before or after. To even out the image, you decide to use a multi-focus scheme that stitches together images taken at three different focal depths into one final image. What is the final frame rate as a result of this change? (c = 1540 m/s, max. depth = 10 cm, 3x spatial compounding) Once again, each frame needs to include 3x more lines, since each line must be imaged at three different focal depths New frame rate = 1 frame 1152lines 1line 0.2 m /1540m/ s 6.7 frames/ s

88 Quantifying image quality 90 How do we quantify image quality and system quality? Two primary metrics (others exist) Signal-to-noise-ratio (SNR): Used to generally quantify system performance as a function of depth Contrast-to-noise ratio (CNR): Used to quantify clarity of a lesion against noise (and speckle)

89 SNR 91

90 SNR 92 Generally will compare a tissue phantom against the same system with transmitters turned off Signal from a tissue phantom Electronic noise

91 SNR 93 Produces a SNR vs. depth curve can quantify when system drops below ideal quality Drops below 6 db From Ultrasound Imaging System Performance, by Ustuner and Holley, 2003 AAPM meeting

92 CNR 94

93 CNR 95 CNR as a function of depth, focusing Focal Depth High Focal Depth Low CNR = 13.2 CNR = 6.5 CNR =8.5 CNR = 8.8 CNR = 2.8 CNR = 6.2

94 New image formation strategies 96 Ultrafast imaging Uses single transmit to cover larger image region Poor resolution, low SNR High max frame rate (>1000 frames per second) Compounding to improve image quality From Cikes et al., Ultrafast Cardiac Ultrasound Imaging JACC Cardiovascular Imaging 2014

95 New image formation strategies 97 Unparalleled temporal resolution and increased amount of information Applications in echocardiography, Doppler, elastography, contrast; new opportunities From Yiu et al., Vector Projectile Imaging UMB 2014 From Mace et al., Functional Ultrasound Imaging of the Brain Nature Methods 2011