www.biomecardio.com – Ultrasound motion imaging – Simulating ultrasound images a very brief introduction by Damien Garcia INSERM researcher, CREATIS, Lyon, France www.biomecardio.com [email protected]April 22, 2019 Disclaimer : The views expressed in this course are those of the author and do not necessarily reflect the multiple positions of the ultrasound community. The examples may contain errors and can carry an implied judgement due to author’s preference for one side of an issue over another. Be critical and take a step back while reading this document!
97
Embed
– Ultrasound motion imaging – Simulating ultrasound images...– Ultrasound motion imaging – Simulating ultrasound images a very brief introduction by Damien Garcia INSERM researcher,
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Disclaimer: The views expressed in this course are those of the author and do not necessarily reflect the multiple positions of the ultrasound community. The examples may contain errors and can carry an implied judgement due to author’s preference for one side of an issue over another.Be critical and take a step back while reading this document!
www.biomecardio.com
TOC
why using simulations? MUST: Matlab UltraSound Toolbox SIMUS: what’s inside?
www.biomecardio.com
why doing simulations?
Before in vitro and in vivo, use computational ultrasound imaging to:
1. test your ultrasound sequences (PW, DW, MLT…)2. optimize your algorithms3. explore multiple configurations
4. compare with others (e.g. challenges)
Computational ultrasound imaging must ideally be:
1. easy to program2. realistic3. easy to parallelize in the 3-D era
The “optimal” methodology (if possible): in silico, in vitro & in vivo
www.biomecardio.com
computational ultrasound imaging
• Computational ultrasound imaging is increasingly used
• Jørgen Jensen, “Field: A program for simulating ultrasound systems.” 1996
2000 2004 2008 2012 2016
50
100
150
200
2018
citations
source: Field II Simulation Program(http://field-ii.dk)
ONLY the diffuse scattering is considered in SIMUS! (as in FieldII)
www.biomecardio.com
Scatterers
Probe
basic principle in SIMUS
www.biomecardio.com
1 cm 1 8 15 22 29 36 43 50 57 64element #
10
20
30
40
50
60
70
time
(μs) after
beamforming
RF signals with SIMUS
www.biomecardio.com
B-mode with SIMUS
ww
w.yale.edu/im
aging/echo_atlas/views/apical_2c.htm
l
www.biomecardio.com
color Doppler with SIMUS
1
2
3
plane wave
propagating downward
128 elements, 5 MHz
(cm
)
17 cm/s
24 cm/s
10 cm/s
source: Shahriari and Garcia.Phys Med Biol, 2018;63:205011.
www.biomecardio.com
color Doppler with SIMUS
source: Shahriari and Garcia.Phys Med Biol, 2018;63:205011.
Doppler
0
17 cm/s
24 cm/s
0 0.5 10
0.2
0.4
(s)
(m/s)
vector Doppler reference (SPH)
www.biomecardio.com
linear acoustic wave equation
Assumptions
1. no dissipative effects (no viscosity, no heat conduction)2. homogeneous, isotropic, elastic medium3. low-amplitude perturbations (small particle velocities, small
fluctuations of pressure and density)
4. ⇒ linearization
𝜕𝜕2𝑝𝑝𝜕𝜕𝑥𝑥2
+𝜕𝜕2𝑝𝑝𝜕𝜕𝑧𝑧2
−1𝑐𝑐2𝜕𝜕2𝑝𝑝𝜕𝜕𝑡𝑡2
= 0
�𝑃𝑃 = ℱ 𝑝𝑝 ⇒𝜕𝜕2 �𝑃𝑃𝜕𝜕𝑥𝑥2
+𝜕𝜕2 �𝑃𝑃𝜕𝜕𝑧𝑧2
+𝜔𝜔2
𝑐𝑐2�𝑃𝑃 = 0
2D acoustic wave equation:
www.biomecardio.com
acoustic field of a 1-D element
1-D element model
1. linear piston in a rigid baffle2. piston vibrating with a uniform normal velocity3. high frequency; far field
4. ⇒ 𝑘𝑘𝑘𝑘 ≫ 1; 𝑘𝑘 ≫ ⁄𝜋𝜋𝑏𝑏2 𝜆𝜆
𝒑𝒑 𝒙𝒙, 𝒛𝒛,𝝎𝝎, 𝒕𝒕
= 𝝆𝝆𝝆𝝆𝒗𝒗𝟎𝟎 𝝎𝝎𝟐𝟐𝒊𝒊𝒊𝒊𝒌𝒌𝒌𝒌 𝑫𝑫𝒌𝒌 𝜽𝜽,𝒌𝒌
𝒆𝒆𝒊𝒊𝒌𝒌𝒊𝒊
𝒌𝒌𝒊𝒊𝒆𝒆−𝒊𝒊𝝎𝝎𝒕𝒕
2𝑏𝑏
𝜃𝜃
𝑘𝑘
𝑘𝑘: wavenumber𝜆𝜆: wavelength
𝑣𝑣0
𝐷𝐷𝑏𝑏 𝜃𝜃, 𝑘𝑘 = sinc 𝑘𝑘𝑏𝑏 sin𝜃𝜃directivity of the element:
www.biomecardio.com
acoustic field of a 1-D array
1-D array
The acoustic field of a 1-D array is the sum of the acoustic fields generated by the single elements
(linear acoustics)
𝒑𝒑 𝒙𝒙, 𝒛𝒛,𝝎𝝎, 𝒕𝒕
= 𝝆𝝆𝝆𝝆𝒗𝒗𝟎𝟎 𝝎𝝎 �𝒏𝒏=𝟏𝟏
𝑵𝑵
𝑾𝑾𝒏𝒏𝒆𝒆𝒊𝒊𝝎𝝎∆𝝉𝝉𝒏𝒏𝟐𝟐𝒊𝒊𝒊𝒊𝒌𝒌𝒌𝒌 𝑫𝑫𝒌𝒌 𝜽𝜽𝒏𝒏,𝒌𝒌
𝒆𝒆𝒊𝒊𝒌𝒌𝒊𝒊𝒏𝒏
𝒌𝒌𝒊𝒊𝒏𝒏𝒆𝒆−𝒊𝒊𝝎𝝎𝒕𝒕
𝜃𝜃1𝑘𝑘1
𝜃𝜃𝑁𝑁𝑘𝑘𝑁𝑁
#1 #2 #3 #N
𝑊𝑊: apodizationΔ𝜏𝜏: delay
#n
𝑘𝑘𝑛𝑛
www.biomecardio.com
receive signals
echo signals
1. The acoustic scatterers become individual monopole point sources when the incident wave reaches them (cylindrical waves in 2-D; spherical waves in 3-D)
2. The scatterers do not acoustically interact with each other (single scattering)
𝒑𝒑𝒒𝒒 𝝎𝝎, 𝒕𝒕
= 𝝆𝝆𝝆𝝆 𝒌𝒌𝒌𝒌𝒗𝒗𝟎𝟎 𝝎𝝎 �𝒎𝒎=𝟏𝟏
# 𝐨𝐨𝐨𝐨 𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩𝐩
𝐁𝐁𝐁𝐁𝐁𝐁𝒎𝒎 �𝒏𝒏=𝟏𝟏
𝑵𝑵
𝑾𝑾𝒏𝒏𝒆𝒆𝒊𝒊𝝎𝝎∆𝝉𝝉𝒏𝒏 𝑫𝑫𝒌𝒌 𝜽𝜽𝒏𝒏𝒎𝒎,𝒌𝒌𝒆𝒆𝒊𝒊𝒌𝒌𝒊𝒊𝒏𝒏𝒎𝒎
⁄𝒊𝒊𝒏𝒏𝒎𝒎 𝒌𝒌𝑫𝑫𝒌𝒌 𝜽𝜽𝒒𝒒𝒎𝒎,𝒌𝒌
𝒆𝒆𝒊𝒊𝒌𝒌𝒊𝒊𝒒𝒒𝒎𝒎
⁄𝒊𝒊𝒒𝒒𝒎𝒎 𝒌𝒌𝒆𝒆−𝒊𝒊𝝎𝝎𝒕𝒕
#q
Tx Rx
𝐵𝐵𝐵𝐵𝐵𝐵:backscattering coefficient
www.biomecardio.com
Field II vs. SIMUS
Time-based frequency-based
Matlab m + mex files Matlab fully open codes
included in MUST
SIMUS
presently, only in 2-D
Field II
1
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES
A synthetic approach based on physical simulators
Olivier Bernard
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES
University of Lyon, France
2
Cardiovascular diseases
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES
Cardiac imaging for diagnosis
►Cardiac imaging
● Assessment of cardiac function (diagnosis / patient follow-up )
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES
4
Cardiac imaging for diagnosis
Strain imaging – echocardiography illustration
Apical 4 chambers
view
Short axis view
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES
5
Cardiac imaging for diagnosis
Longitudinal Radial Circumferential
L L L
𝒆 = ∆𝑳
𝑳
Normalized deformation
Strain imaging – echocardiography illustration
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES
6
Echocardiography illustration
Cardiac imaging for diagnosis
Longitudinal strain
Source: GE Healthcare web site
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES
7
Cardiac imaging for diagnosis
Echocardiography illustration
Radial strain
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES
Source: GE Healthcare web site
8
Cardiac imaging for diagnosis
Echocardiography illustration
Circumferential strain
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES
Source: GE Healthcare web site
9
Cardiac imaging for diagnosis
►Sensitive to change of systolic function
● Strong potential for detecting heart diseases at early stage
► Ischemic case: reduced motion of specific segments
Myocardial strain
LCX: Occlusion of Left Circumflex RCA: Occlusion of Right Coronary Artery LADdist: Distal occlusion of the Left Anterior Descending Artery LADprox: Proximal occlusion of the Left Anterior Descending Artery
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES
10
Cardiac imaging for diagnosis
Myocardial strain
► LCX example
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES
11
Cardiac imaging for diagnosis
Myocardial strain
►So everything is beautiful in a wonderful word ?
● Not really…
►Only global longitudinal strain (GLS) is used (in ultrasound)
►Regional strain NOT used (despite the clinical interests)
Strain measurements are not reproducible enough Needs for automatic and reproducible measurements
Solid quantitative validations are required
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES
12
Cardiac imaging for diagnosis
Validation of cardiac strain quantification
Manual tracking Physical
phantom
Animal experiments
Realistic synthetic images
Straightforward Real
acquisitions Measure strain
directly Dense strain Ground-truth
• Tedious • Inter and
intra-expert variability
• Realism (image quality/involved structures) not yet sufficient
• Image quality is too good
• Ethical question
• Let’s see what’s going on
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES
13
Generation of realistic synthetic images
Motivations
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES
14
Motivations
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES
15
● Physical principle
● Physical simulator
● Proposed pipeline
Ultrasound modality
Generation of realistic synthetic images
Ultrasound modality
GENERATION OF REALISTIC SYNTHETIC ULTRASOUND IMAGES