SYSTEMS AND METHODS FOR COMPOSITE MYOCARDIAL ELASTOGRAPHY

Abstract

Systems and methods for composite myocardial elastography are presented. In exemplary embodiments of the present invention an imaging modality field of view, such as, for example, that of ultrasound, can be divided into N sectors, each having 1/Nth of a full field of view. In exemplary embodiments of the present invention a temporal series of 2D ultrasound images for each of the N sectors can be acquired over a duration of one or more periods of a periodic signal. Substantially simultaneously, such a periodic signal can also be acquired, wherein each of said series of 2D ultrasound images for each sector can be triggered using said periodic signal. For example, for ultrasound imaging of the heart, an ECG signal can function as such a periodic signal. The data from the various N sectors can be synchronized in time using the ECG signals, and the ultrasound signals from each of the N sectors combined to generate a series of composite ultrasound images at the frame rate of the individual sectors. In exemplary embodiments of the present invention such a composite image can be further processed to estimate displacement between consecutive frames, remove noise, accumulate displacement with time for an entire cardiac cycle, and derive strain in the cardiac muscle, vessel or any other organ or tissue under motion. In exemplary embodiments of the present invention the derived strain data can be overlaid onto all or part of the composite ultrasound images, and one or more of such overlaid images can be displayed to a user.

Description

INCORPORATION-BY-REFERENCE OF A COMPUTER PROGRAM LISTING APPENDIX

A computer program listing appendix is included pursuant to 37 C.F.R. 1.52(e) and is hereby incorporated by reference in its entirety. The computer program listing appendix was submitted via EFS on Jul. 13, 2012. The computer program listing appendix includes the following 14 files: a table of contents, submitted as the ASCII text file toc.txt, is 242 bytes; analyzeNrf.m, submitted as the ASCII text file analyzeNrf_m.txt, is 8,762 bytes; cutECG.m, submitted as the ASCII text file cutECG_m.txt, is 1,634 bytes; findSectorOverlap.m, submitted as the ASCII text file findSectorOverlap_m.txt, is 3,714 bytes; matchbestECG.m, submitted as the ASCII text file matchbestECG_m.txt, is 4,770 bytes; data2rgb.m, submitted as the ASCII text file data2rgb_m.txt, is 689 bytes; initOverlay.m, submitted as the ASCII text file initOverlay_m.txt, is 1,967 bytes; overlayData.m, submitted as the ASCII text file overlayData_m.txt, is 4,141 bytes; overlayimage.m, submitted as the ASCII text file overlayimage_m.txt, is 662 bytes; tiss2rgb.m, submitted as the ASCII text file tiss2rgb_m.txt, is 135 bytes; makeCardiacMovie.mat, submitted as the ASCII text file makeCardiacMovie_mat.txt, is 5,588 bytes; getPolTransformMap.m, submitted as the ASCII text file getPolTransformMap_m.txt, is 3,144 bytes; appPolTransform.m, submitted as the ASCII text file appPolTransform_m.txt, is 1,803 bytes; getmyparams.m, submitted as the ASCII text file getmyparams_m.txt, is 419 bytes. Each file included in the computer program listing appendix described above was created on Sep. 23, 2011. The computer program listing appendix does not include any new matter which goes beyond the disclosure of the application as filed.

TECHNICAL FIELD

The present invention relates to medical imaging, and in particular to increasing the frame rate of ultrasound imaging by dividing the field of view into sectors, obtaining a series of ultrasound images for each sector, synchronizing the images and combining them to form a composite high-frame rate image.

BACKGROUND

Ultrasound imaging can be a useful tool in cardiology, such as, for example, in the diagnosis of myocardial infarctions. Ultrasound imaging of the heart, known as echocardiography, can be used, for example, to derive strains, which are related to the contractility of the heart muscle. However, current methods of real-time raw data plane in a given system, which can be defined by a spanned angle (i.e., arc length according to the center of an imaging probe) and a chosen depth (beam direction). Tracking small regions of the heart at frame rates of 50 fps is difficult. Moreover, because strains involve motion of the heart muscle, a frame rate (effectively a sampling rate of the displacement function over time) is required to be high enough such that interframe motion is relatively small to be accurately estimated. Using conventional frame rates, strain image results tend to be both noisy and unreliable This is because the lower the frame rate, the less correlated any two consecutive frames are, which makes radio-frequency (RF)-cross-correlation based motion estimation techniques less accurate. One quantitative measure of the noise on strain images is the elastographic signal-to-noise ratio, or SNRe.

With the high frame-rate composite imaging, precise and detailed motion/strain estimates in full view can be obtained and further used to differentiate abnormal from the normal myocardium and even detect the onset and extent of the diseased muscle. From theoretical and in vivo examples, the difference between strain in a normal and an abnormal myocardium can be large in the case of acute infarction but also subtle in the case of chronic infarction, infarction scars or small infarcted regions. Visualization in the latter case is more challenging and an imaging modality that estimates the strain at high precision and thus SNRe is warranted. RF-based tracking can provide such precision to estimate subtle motion changes in the pathological myocardium. Most importantly, the ischemic region will undergo abnormal, i.e., smaller or reverse, motion due to its reduced contractility. Estimation of the resulting smaller motion and/or strain (compared to the normal case) also requires higher precision of the method used. Again, RF-based tracking (as opposed to the faster and more commonly used B-mode tracking) will yield the highest precision estimate and thus highest quality images. Due to the higher sensitivity of RF-based tracking, i.e., the higher decorrelation rate, RF tracking is best used at the highest frame rates, where consecutively acquired RF echoes are best matched because they are recorded at small incremental time intervals.

In a similar way, the same invention can be applied for visualization of all transient motion effects in tissues or vessels, such as the pulse wave traveling in the arterial tree at each heartbeat, respiratory motion, or the pulsation of internal vessels in organs, such as the liver, pancreas, kidney, thyroid or prostate.

What is thus needed in the art are systems and methods that can increase the ultrasound frame rate so as to be sufficiently high to capture cardiac motion and provide meaningful strain image results.

SUMMARY

Systems and methods for composite myocardial elastography are presented. In exemplary embodiments of the present invention an imaging modality field of view, such as, for example, that of ultrasound, can be divided into N sectors, each having 1/Nth of a full field of view. In exemplary embodiments of the present invention a temporal series of 2D ultrasound images for each of the N sectors can be acquired over a duration of one or more periods of a periodic signal. Substantially simultaneously, such a periodic signal can also be acquired, wherein each of said series of 2D ultrasound images for each sector can be triggered or gated using said periodic signal. For example, for ultrasound imaging of the heart, an ECG signal can function as such a periodic signal. The data from the various N sectors can be synchronized in time using the ECG signals, and the ultrasound signals from each of the N sectors combined to generate a series of composite ultrasound images at the frame rate of the individual sectors. In exemplary embodiments of the present invention such a composite image can be further processed to estimate displacements between consecutive frames, remove noise, accumulate displacements with time for an entire cardiac cycle, and derive strain in the cardiac muscle. In exemplary embodiments of the present invention the derived strain data can be overlaid onto all or part of the composite ultrasound images, and one or more of such overlaid images can be displayed to a user.

BRIEF DESCRIPTION OF THE DRAWINGS

The patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.

FIG. 1 illustrates exemplary sector data acquisition according to an exemplary embodiment of the present invention;

FIG. 2 illustrates exemplary ECG-gated sector data acquisition according to an exemplary embodiment of the present invention;

FIG. 3 depicts an exemplary overall process flow according to an exemplary embodiment of the present invention;

FIGS. 4( a)-14(b) depict the intermediary outputs of various sub-processes of the exemplary process flow of FIG. 3 according to an exemplary embodiment of the present invention;

FIGS. 12( e)-12(l) and 13(e)-13(l) depict the underlying image and the overlay separately, in grayscale, and correspond to FIGS. 12(a)-12(d) and 13(a)-13(d), respectively;

FIG. 15 depicts an exemplary B-mode image obtained according to an exemplary embodiment of the present invention;

FIGS. 16( a)-16(n) depict incremental axial displacements during systole at 50 frames per second according to an exemplary embodiment of the present invention;

FIGS. 16( o)-16(ap) are grayscale images corresponding to FIGS. 16(a)-16(n) which show the displacement separately from the B-mode image;

FIGS. 17( a)-17(k) depict incremental axial displacement during diastole at 50 frames per second according to an exemplary embodiment of the present invention;

FIGS. 17( l)-17(ag) are grayscale images corresponding to FIGS. 17(a)-17(k) which show the displacement separately from the B-mode image;

FIGS. 18( a)-18(n) depict incremental strain images during systole at 50 frames per second according to an exemplary embodiment of the present invention;

FIGS. 18( o)-18(ap) are grayscale images corresponding to FIGS. 18(a)-18(n) which show the strain separately from the B-mode image;

FIGS. 19( a)-19(k) depict incremental strain images during diastole at 50 frames per second according to an exemplary embodiment of the present invention;

FIGS. 19( l)-19(ag) are grayscale images corresponding to FIGS. 19(a)-19(k) which show the strain separately from the B-mode image;

FIGS. 20( a)-20(n) depict incremental axial displacement during systole at 136 frames per second according to an exemplary embodiment of the present invention;

FIGS. 20( o)-20(ap) are grayscale images corresponding to FIGS. 20(a)-20(n) which show the displacement separately from the B-mode image;

FIGS. 21( a)-21(n) depict incremental axial displacement during diastole at 136 frames per second according to an exemplary embodiment of the present invention;

FIGS. 21( o)-21(ap) are grayscale images corresponding to FIGS. 21(a)-21(n) which show the displacement separately from the B-mode image;

FIGS. 22( a)-22(n) depict incremental strain images during systole at 136 frames per second according to an exemplary embodiment of the present invention;

FIGS. 22( o)-22(ap) are grayscale images corresponding to FIGS. 22(a)-22(n) which show the strain separately from the B-mode image;

FIGS. 23( a)-22(n) depict incremental strain images during diastole at 136 frames per second according to an exemplary embodiment of the present invention;

FIGS. 23( o)-23(ap) are grayscale images corresponding to FIGS. 23(a)-23(n) which show the strain separately from the B-mode image;

FIG. 24 depicts exemplary process flow charts for an exemplary implementation of the present invention on a programmable ultrasound machine;

FIG. 25 illustrates an exemplary multi-sector combination technique for high frame rate full view ultrasound according to an exemplary embodiment of the present invention;

FIG. 26 illustrates irregular ECG interpolation according to an exemplary embodiment of the present invention;

FIGS. 27( a)-27(b) depict a comparison of image quality before and after overlap processing according to an exemplary embodiment of the present invention;

FIGS. 28( a)-28(b) depict a comparison of 7 sector B-mode composite images of an exemplary heart's long axis view according to an exemplary embodiment of the present invention with 100% B-mode images without comparison;

FIGS. 29( a)-29(l) depict propagation of an exemplary EM wave propagating from apex to base during late diastole captured using imaging techniques according to an exemplary embodiment of the present invention;

FIGS. 29( m)-29(aj) are grayscale images corresponding to FIGS. 29(a)-29(l) which show the displacement separately from the B-mode image;

FIGS. 30( a)-30(l) depict propagation of an exemplary EM wave propagating from septum to posterior wall during diastole captured using imaging techniques according to an exemplary embodiment of the present invention;

FIGS. 30( m)-30(ah) are grayscale images corresponding to FIGS. 30(a)-30(l) which show the displacement separately from the B-mode image;

FIGS. 31( a)-31(l) depict another propagation of an exemplary EM wave captured using imaging techniques according to an exemplary embodiment of the present invention;

It is noted that the patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawings will be provided by the U.S. Patent Office upon request and payment of the necessary fee.

It is also noted that certain figures are described as being in color. For the PCT counterpart to this application these images are provided for completeness, but are shown in grayscale to comply with PCT rules and regulations. The color scales which appear on the right of images thus refer to the color overlays onto B-mode images. All color overlays are provided separately in the grayscale images as well.

DETAILED DESCRIPTION

The present invention involves increasing the effective frame rate of an imaging modality by utilizing a periodic signal to synchronize various stored smaller portions of full scan frames and combine them in a temporally correct manner. In exemplary embodiments of the present invention the imaging modality can be, for example, ultrasound, the periodic signal can be, for example, an electrocardiogram, and the imaged anatomical area can be, for example, the heart, such imaging results being used, in particular, to analyze cardiac muscle strain.

In exemplary embodiments of the present invention, a technique known as Composite Myocardial Elastography (CME) can be performed to increase the frame rate of conventional cardiac ultrasound imaging. This technique utilizes ECG-gating to acquire several narrow views (sectors) at high frame rates (e.g., 136 fps) over several cardiac cycles. In exemplary embodiments of the present invention an ECG-triggered elastocardiographic method can include, for example, combining displacements and strains obtained at smaller fields-of-view and aligning them based on a simultaneously acquired ECG signal and spatial information into a final composite image. Such a composite image can thus have a full field of view, a significantly higher frame rate, and a significantly higher SNR_e. The higher the frame rate, the better correlated the consecutive frames are and therefore, the higher the correlation coefficient used in estimating the displacement. This translates to a higher quality displacements and higher SNR of the elastograms, i.e., higher SNR_e. The SNR_e is directly proportional to the correlation coefficient ρ according to the Cramer-Rao Lower Bound (CRLB) is

${\mathrm{SNR}}_e=\frac{\pi \stackrel{\_}{\hat{\varepsilon }}T\sqrt{\left(B^3+12{\mathrm{Bf}}_0^2\right)\Delta t}}{\sqrt{12\left[\frac{1}{{\rho }^2}\left(1+\frac{1}{{\mathrm{SNR}}_s^2}\right)-1\right]}},$

where σ=ρ_xσ_yρ_z is the correlation coefficient (equal to the product of the correlation coefficients associated with motion in each direction), SNR_s is the sonographic signal-to-noise ratio, B is the bandwidth and f₀ is the frequency (Varghese and Ophir 1997). The frame rate FR is equal to c/(2(N′/N)D), where c is the speed of sound, N′ is the number of RF signals, N is the number of sectors and D is the depth. The speed of sound in soft tissues is equal to 1540 m/s and the depth in human echocardiography is typically 10-12 cm. The highest obtainable frame rate will be achieved when the smallest number of RF signals (or, largest number of sectors) is used, i.e., N′=1. In that case, the frame rate would be equal to 6250 frames/s (or, 6.25 kHz). When N′=128 (i.e., the conventional number of RF signals used), the lowest frame rate will be achieved, i.e., 48.83 frames/s (or, 0.049 kHz). For N={6, 12, 20}, the corresponding frame rates are FR={1.04, 0.52, 0.26} kHz.

Thus, in exemplary embodiments of the present invention, a full field image can be reconstructed by combining the data from such sectors, or narrow views, into a combined image. The combined image will thus have the high frame rate of the individual sectors from which it is composed.

Strain images produced at such a higher frame rate are less-affected by noise relative to lower frame rates. This is because a higher cross-correlation coefficient can be obtained at higher frame rates. Systems and methods according to exemplary embodiments of the present invention can be utilized in various analytic, diagnostic and therapeutic applications. One exemplary application is, for example, detecting and quantifying the extent of ischemia and infraction in the myocardium at and beyond its onset due to the associated significantly alerted stiffness of the muscle.

As noted above, conventional real-time raw data acquisition of a full view of the heart limits the data acquisition rate to approximately 50 frames per second (fps). At this frame rate, tracking of smaller regions of the heart is difficult. In general, transmural (or, across the myocardial wall) infarcts can be minute (the wall maximal thickness is 1 cm and still cause problematic cardiac function that may not be evident through the use of ECG or enzymic activity (i.e., blood tests). High resolution and high sensitivity imaging will allow early detection and potentially serve as a screening technique. Additionally, strain image results produced at this rate are noisy and not reliable. Elastography involves motion imaging. Like photography, if the shutter speed is too low, the resulting picture of a moving subject will be blurry. However, if the shutter speed is high enough, the picture of a moving subject will be clear. The same principle applies in elastography that attempts to obtain motion estimates and images of the heart during contraction of the latter. Exemplary embodiments of the present invention can be used to solve such problems of conventional ultrasound systems. Next described are exemplary embodiments according to the present invention with reference to the figures.

In exemplary embodiments of the present invention, instead of having an ultrasound probe fire all of its transducer elements sequentially to complete one frame, then return to the first transducer element and repeat the process to produce a second frame, the series of sensors can be divided into a number, N, of sectors. For each of the N sectors, a number, M, of frames per second can be obtained by firing only the ultrasound emitters/detectors associated with that particular sector. Thus, for each sector, M frames of imaging data can be obtained per second. By imaging each of the N sectors sequentially, and synchronically combining the data associated with each of the N sectors together into a full field of view composite image, the equivalent of M frames per second of a full view image can be obtained, at a tradeoff of the data from the various sectors not being acquired simultaneously. For example, where each sector data acquisition takes one cardiac cycle, data from Q cardiac cycles can be combined to form a composite image. This process results in a significantly higher data rate, equal to M frames per second as opposed to M/N frames per second for conventional full view ultrasound imaging. This increased frame rate is due to the fact that only a subset of the available transducers on the ultrasound probe need to fire for each frame. The tradeoff is that each acquired frame is considerably narrower than a full view, and these narrow frames must be somehow synchronically combined into a set of full view frames.

Thus, the key to combining various narrow view images respectively corresponding to each of the N sectors into a combined full view is the ability to synchronize the various sectors in time and in space. In exemplary embodiments of the present invention such synchronization can be accomplished, for example, by triggering data sector acquisition using a periodic signal, such as, for example, an electrocardiogram signal (ECG), which can also be stored and later used to temporally synchronize the various data sectors. Thus, assuming that the ECG signal is sufficiently auto-correlated across cardiac cycles, such composite images are equivalent to a full field of view image at the higher frame rate of the narrow sectors.

Sector data acquisition is illustrated in FIG. 1. In sector data acquisition each raw data sector can contain, for example, a series of 2D images over a defined period of time, such as, for example one or more cardiac cycles. As noted, each sector contains 1/N of the area of a full field of view. FIG. 2 illustrates ECG-gated data acquisition. In the exemplary embodiment of FIG. 2 the ultrasound sensors have been divided into six (6) different sectors. These constitute raw data sectors 210. Each sector can be synchronized relative to the beginning of a simultaneously acquired ECG signal 220, by, for example, identifying the QRS peak using known techniques. The ECG signal can then be used to synchronize the various data sectors in time, such as, for example, by having each sector begin at a defined time interval at or after the QRS peak.

Thus, for example, using an ultrasound probe that has 128 emitter/detector elements, sector 1 can be assigned to elements 1-21, sector 2 can assigned to elements 22-42, sector 3 can be assigned to elements 43-63, sector 4 can be assigned to elements 64-84, sector 5 can be assigned to elements 85-105 and sector six can be assigned to elements 106-126. For each sector, a series of images (frames) can be taken over one or more cardiac cycles, for example, and the data from each sector can be combined to make a set of full view ultrasound images over the same time duration.

Moreover, because while the ECG signal is highly auto-correlated in time but still may have variations from one cycle to the next, sector data from a number of different cardiac cycles can, for example, be obtained, and the best matches for each sector can be combined into a composite series of ultrasound images representing one best-fit “composite” cardiac cycle. This can be done, for example, by picking one cardiac cycle's worth of data for each sector from the multiple cardiac cycle data acquired for each sector. For example, the best matching segments of ECG based on length and shape can be used in such a calculation. As noted above, this is possible because the images for each sector can be acquired over a time span of multiple (for example, three) cardiac cycles. This “mix and match” sector combination can be done, for example, using the exemplary “matchbestECG.m” program provided in the computer program listing appendix. FIG. 3 depicts an overall process flow according to an exemplary embodiment of the present invention. Beginning at the top of FIG. 3 there can be seen sectors 1 through N (in the depicted example N=6), where for each sector raw data 305 and simultaneously acquired ECG data 310 can be collected. Using some matching criteria, such as, for example, the correlation coefficient obtained from matching, for example, the R-wave peaks of the ECG signal, as described above, at 320 the best one ECG cycle worth of data for each sector can be chosen and, at 330 the sector raw data from the various sectors can be combined using spatial (angle, depth) and temporal (ECG signal) parameters. Moreover, the sectors can have an overlap of, for example, 10-20% for registration purposes. The starting depth (i.e., beam direction), starting angle and angular increment (i.e., azimuthal direction) of each sector are recorded and used to combine multiple sectors. For example, the first sector can span from 0-11 degrees, the second 9-21, the third 19-31, etc. The spatial registration can thus be performed using the angle and depth information and temporal registration can be obtained using the ECG signal. After combination at 330, the result is a series of composite images 325 such as is shown, for example, in FIG. 15. The B-Mode image of FIG. 15 is a standard full view ultrasound image, and the series obtained in this manner has a higher temporal sampling rate than otherwise possible with full field of view imaging. The remaining processes depicted in FIG. 3 relate to displacement identification and strain derivation therefrom. Thus, at 340, the displacements can be estimated, for example, between consecutive time frames. Here, displacements are estimated for the entire composite FOV. At 350, for example, a noise removal algorithm can be implemented, such as, for example, the “cleannoise2.m” program provided in the computer program listing appendix. Noise removal utilizes the information of correlation coefficients. Only the estimates with high correlation coefficients above 0.7 are deemed reliable. Those with lower correlation coefficients will be replaced by the average of the surrounding estimated values. At 360, for example, the displacement can be accumulated with time so as to track motion for an entire cardiac cycle, for example.

Continuing with reference to FIG. 3, at 365 strain in the cardiac muscle can, for example, be derived from the accumulated displacement generated at 360. Strain can be defined in terms of the gradient of the displacement. The cumulative 2D displacement 360 can thus, for example, be written as u=u_xe_x+u_ye_y, where u_x and u_y are lateral and axial displacements, respectively, and e_x and e_y are unit coordinate base vectors in lateral and axial directions, respectively. The cumulative 2D displacement gradient tensor, ∇u, can be defined as:

$\begin{array}{cc}\nabla \underset{\_}{u}=\left[\begin{array}{cc}\frac{\partial u_x}{\partial x}& \frac{\partial u_x}{\partial y}\\ \frac{\partial u_y}{\partial x}& \frac{\partial u_y}{\partial y}\end{array}\right],& \left(3\right)\end{array}$

and the strain tensor, E, can be defined as

$\begin{array}{cc}E=\frac{1}{2}\left(\nabla \underset{\_}{u}+{\left(\nabla \underset{\_}{u}\right)}^T+{\left(\nabla \underset{\_}{u}\right)}^T\nabla \underset{\_}{u}\right),& \left(4\right),\end{array}$

where (∇u)^T is the transpose of ∇u. The lateral and axial strains are the diagonal components of E, i.e., E_xx, and E_yy, respectively.

Given such derived strain, at 370, for example, the derived strain can be overlaid onto B-Mode image 325 generated at 330. Finally, this overlay can be displayed on the screen, as shown in FIGS. 18(a)-19(ag) and 22(a)-23(ap), and process flow ends.

Although in the above process flow description reference has only been made to a few of the exemplary Matlab™ programs provided in the computer program listing appendix, those skilled in the art will understand how each of those exemplary programs can be used to implement the various processes illustrated in FIG. 3. The following table correlates the various exemplary processes of FIG. 3 with either exemplary Matlab™ programs provided in the computer program listing appendix, along with the inputs and outputs to such programs, or with pseudocode (for steps 340, 350, 360 and 365), as provided immediately after the table. For the exemplary Matlab™ code the following variables are used:

• • M=number of sectors;
  • N=length of ECG signals for three cardiac cycles;
  • N1=length of ECG signals for the cardiac cycle with maximum length;
  • Nf=number of total frames in the beginning; and
  • nFrames=number of frames per sector per cardiac cycle.

	Function name	Input	Output
Main	analyzeNrf	N/A	N/A
script	Reconstruct several
	overlapping small sector
	RF acquisitions.
305	OPENCLP	filename	fid
	(provided by EchoPAC)
	Open the DICOM format
	files saved in GE Vivid 5
	RDTISSUE	Filename	alltiss (3D matrix)
	(provided by EchoPAC)	infoTISS (1-by-Nf)
	Read the b-mode data	startframe (2*M-by-2)
		stopframe (2*M-by-2)
	Readiq	Filename	tiq (2D matrix)
	(provided by EchoPAC)	infoIQ (1-by-Nf)
	Read the in-phase	startframe (2*M-by-2)
	quadrature (IQ) data (i.e.,	nFrames (2*M-by-2)
	raw data)
	IQ2RF	tiq	allrf (3D matrix)
	(provided by EchoPAC)	infoIQ
	Convert IQ data to Radial-
	frequency (RF) data
310	readecg	Filename	ecgTISS (1-by-N)
	(provided by EchoPAC)	infoTISS (1-by-Nf)	syncTISS (1-by-N)
	Read the ECG data stored
	in GE Vivid 5
320	matchbestECG	ecgTISS (1-by-M cell)	ecgbest (M-by-2)
	Finds the best matching		{ecgTISS1, ecgTISS2,
	segments of ECG based on		ecgTISS3, ecgTISS4,
	length and shape		ecgTISS5} (1-by-M cell)
	cutECG	ECGfull (1-by-N)	ECGparts (3-by-N1)
	This function cuts up an		ECGtrigger (1-by-N1)
	ECG signal into N parts.		ECGpeakshift (1-by-M)
		syncTISS	startframe (2*M-by-2)
		syncIQ	stopframe (2*M-by-2)
		ecgbest (2*M-by-2)
325	makeCardiacMovie	overlayAll	Movie
	The main script to make	ecgIQ1
	the movie displaying	syncIQ1
	overlaid displacement	parameters
	image	scanconvert (true or false)
	getPolTransformMap	prpol (2-by-4)	mrows (2D matrix)
	Computes a polar-to-	prcart (2-by-4)	mcols (2D matrix)
	cartesian coordinate		mmask (2D matrix)
	transformation map defined
	by prpol and prcart.
	appPolTransform	Data	overlaysc (4D matrix)
	applies a polar-to-cartesian	mrows (2D matrix)
	coordinate transformation	mcols (2D matrix)
	to imgpol.	mmask (2D matrix)
		flag = 1 (linear
		interpolation)
330	readinfo	fid	tinfo
	(provided by EchoPAC)	0
	Read all EchoPAC file
	information
	getmyparams	tinfo	tparams->
	get parameters from each	paramnamesTISS or	paramsIQ (M-by-6)
	file	paramnamesIQ	paramsTISS (M-by-6)
	findSectorOverlap	paramsIQ or paramsTISS	sectorBMbeams (M-by-8)
	Utilizes params matrix to	(M-by-6)	or sectorIQbeams (M-by-
	find overlapping regions of	centerAngle	8)
	sectors, and returns	iSA = 3
	sectorBeams, a matrix	iAU = 4
	containing information	iNB = 6
	about overlap and which	iSD = 1
	data to utilize from each	iDU = 2
	sector for the reconstructed	iNS = 5
	whole sector.
340	See pseudocode below
350	See pseudocode below
360	See pseudocode below
365	See pseudocode below
370	initOverlay.m	paramsIQ (2D matrix)	oparams (structure)
		paramsTISS (2D matrix)
	overlayData.m	tiss (3D matrix)	overlayAll (4D matrix)
		data (3D matrix)
		oparams (structure)
		TISSISRF (flag)
	overlayimage.m	tissRGB (3D matrix)	overlay(3D matrix)
		dataRGB (3D matrix)
		mask (3D matrix)
	tiss2rgb.m	tiss (2D matrix)	tissRGB (3D matrix)
		paramsTISS (2D matrix)
		DYN (scalar)
		GAIN (scalar)
	data2rgb.m	dataResize (2D matrix)	dataRGB (3D matrix)
		CMAP
		dataLIMS (1-by 2-vector)

Pseudocode referred to in above Table for 340, 350, 360 and 365 of FIG. 3

340

Read raw data

Initialize window_size, overlap, interp_factor

Set sample_count to (1−overlap)*window_size

FOR 1: shift: total_sample_points

• • FOR lateral_beam_count
    • 2D interpolation of RF signals
    • Calculate cross-correlation coefficient between signals
    • Find the maximal coefficient
    • Cosine interpolation around the maximal coefficient
    • Store the location with the interpolated maximal coefficient
    • Calculate lateral and axial displacements based the store location
  • ENDFOR

ENDFOR

Generate RF signals with the removal the axial displacement and recalculate the lateral displacement

lateral_disp=lateral displacement after axial displacement correction

axial_disp=the initial estimated axial displacement

Return lateral_disp, axial_disp, cross-correlation_coefficient

350

Function CLEAN_NOISE (displacement, cross_correlation_coefficients)

FOR each displacement

IF cross_correlation_coefficient<threshold

• • The displacement value is updated with the average of the neighboring values

ENDIF

ENDFOR

360

function [cum_lateral_disp, cum_axial_disp]=CUM_DISP (lateral_disp, axial_disp)

365

function strain_tensor_—2D=CALC_STRAIN (lateral_disp, axial_disp)

G11=the gradient of lateral_disp along the lateral direction

G12=the gradient of lateral_disp along the axial direction

G21=the gradient of axial_disp along the lateral direction

G22=the gradient of axial_disp along the axial direction

G(1, 1)=G11; G(1, 2)=G12; G(2, 1)=G21; G(2, 2)=G22;

strain_tensor_—2D=½*(G+transpose (G)+transpose (G)*G)

Return strain_tesnsor_—2D

FIGS. 4( a)-14(b) depict exemplary intermediate outputs of various exemplary sub-processes of FIG. 3, generated using the exemplary Matlab™ source code provided in the computer program listing appendix. These intermediate results, and how the various modules in the exemplary source code can be used to generate them, are next described with reference to FIGS. 4(a)-14(b). It is noted that although in connection with FIG. 3, the number of sectors N was, for example, six, in the exemplary images of FIGS. 4(a)-4(e) and 5(a)-5(e) only five sectors were used. In general, the number of sectors depend on the individual sector size selected and on the size of the left ventricle imaged.

FIGS. 4( a)-4(e) depict five exemplary sector outputs. The sectors were used to image the whole long axis view of the left ventricle. Each sector shows the raw data. FIGS. 5(a)-5(e) show an exemplary 3 cardiac cycles (ECG signals) obtained while each sector was being imaged. From this data, FIGS. 6(a)-6(e) show the best match cardiac cycle (ECG signal) from each sector. The best matched cardiac cycle can be determined according to the highest cross-correlation coefficient obtained.

FIG. 7 shows the combined raw data of the long axis view of the left ventricle, FIGS. 8(a)-8(b) show incremental lateral and axial displacements before noise removal, and FIGS. 9(a)-9(b) show incremental lateral and axial displacements after noise removal.

FIGS. 10( a)-10(b) show cumulative lateral and axial displacements from end-diastole (ED) to end-systole (ES). FIGS. 11(a)-11(b) show cumulative lateral and axial strains from ED to ES. FIGS. 12(a)-12(d) depict exemplary cumulative (a) lateral and (c) axial displacements from tagged MRI (tMRI) imaging between end-diastole (ED) and end-systole (ES), respectively; and cumulative (b) lateral and (d) axial displacements from 2D myocardial elastography (2DME) between ED and ES, respectively. All the depicted short-axis images were acquired approximately at the papillary muscle level and shown at end-systolic configuration. Similarly, FIGS. 13(a)-13(d) depict cumulative (a) lateral and (c) axial systolic strains from tMRI between ED and ES, respectively, and cumulative (b) lateral and (d) axial systolic strains from 2DME between ED and ES, respectively. All the short-axis images were acquired approximately at the papillary muscle level and shown at end-systolic configuration.

It is noted that FIGS. 12(e)-12(l) and 13(e)-13(l) depict the underlying image and the overlay separately, in grayscale, and correspond to FIGS. 12(a)-12(d) and 13(a)-13(d), respectively.

Finally, FIGS. 14(a)-14(b) show an exemplary B-mode long axis view of the left ventricle before and after scan conversion.

As noted above, FIG. 15 is an exemplary B-Mode image such as, for example, that generated at 325 in FIG. 3 from the combined sector data.

Exemplary B-Mode Images

1. Low Frame Rate Displacement Images for Systole and Diastole—FIGS. 16(a)-17(ag)

FIGS. 16( a)-16(n) and 17(a)-17(k) illustrate exemplary displacement results obtained during systole (contraction) and diastole (expansion), respectively, for a series of time points, wherein displacement has been color coded in each image according to the color coded bar key appearing at the right of each image. The time point within the cardiac cycle at which each image has been acquired is indicated by the solid ball below each image. FIGS. 16(a)-16(n) and 17(a)-17(k) are conventional full view images acquired in a conventional manner; they thus represent a frame rate of 50 frames per second. The displacement is overlayed in color on the B-mode images. FIGS. 16(o)-16(ap) and 17(l)-17(ag) are greyscale images corresponding to FIGS. 16(a)-16(n) and 17(a)-17(k), respectively. In the greyscale images of FIGS. 16(o)-16(ap) and 17(l)-17(ag), the displacement has been separated from the B-mode images for ease of viewing. Thus each image in FIGS. 16(a)-16(n) and 17(a)-17(k) corresponds to two images in FIGS. 16(o)-16(ap) and 17(l)-17(ag)—one for the B-mode image, the other for the displacement. For example, FIG. 16(a) corresponds to FIGS. 16(o) and 16(p); FIG. 16(o) presenting the B-mode image, and FIG. 16(p) the overlay. A similar correspondence exists in FIGS. 17(a)-23(ap), described below.

2. Low Frame Rate Strain Images for Systole and Diastole—FIGS. 18(a)-19(ag)

Similarly, FIGS. 18(a)-18(n) and 19(a)-19(k) show a series of strain images obtained during systole (contraction) and diastole (expansion), respectively, for a series of time points, wherein the strain has been color coded in each image according to the color coded bar key appearing at the right of each image. The time point within the cardiac cycle at which each image has been acquired is indicated by the solid ball below each image. FIGS. 18(a)-18(n) and 19(a)-19(k) are conventional full view images acquired in a conventional manner; they thus represent a frame rate of 50 frames per second. The strain is overlayed in color on the B-mode images. FIGS. 18(o)-18(ap) and 19(l)-19(ag) are greyscale images corresponding to FIGS. 18(a)-18(n) and 19(a)-19(k), respectively. In the greyscale images of FIGS. 18(o)-18(ap) and 19(l)-19(ag), the displacement has been separated from the B-mode images for ease of viewing. Thus each image in FIGS. 18(a)-18(n) and 19(a)-19(k) corresponds to two images in FIGS. 18(o)-18(ap) and 19(l)-19(ag)—one for the B-mode image, the other for the displacement.

3. High Frame Rate Displacement Images for Systole and Diastole—FIGS. 20(a)-21(ap)

Similarly, FIGS. 20(a)-20(n) and 21(a)-21(k) show a series of composite displacement images made by combining various sectors according to an exemplary embodiment of the present invention, thus obtaining a high frame rate of, for example, 136 frames per second. The time point within the cardiac cycle at which each image has been acquired is indicated by the solid ball below each image. As above for the low frame rate images, the displacement is overlayed in color on the B-mode images.

Similarly, FIGS. 20(o)-20(ap) and 21(l)-21(ag) are grayscale images corresponding to FIGS. 20(a)-20(n) and 21(a)-21(k), respectively. In the grayscale images of FIGS. 20(o)-20(ap) and 21(l)-21(ag), the displacement has been separated from the B-mode images for ease of viewing. Thus each image in FIGS. 20(a)-20(n) and 21(a)-21(k) corresponds to two images in each of FIGS. 20(o)-20(ap) and 21(l)-21(ag)—one for the B-mode image, the other for the displacement.

4. High Frame Rate Strain Images for Systole and Diastole—FIGS. 22(a)-23(ap)

Similarly, FIGS. 22(a)-22(n) and 23(a)-23(n) show a series of composite strain images made by combining various sectors according to an exemplary embodiment of the present invention, thus obtaining a high frame rate of, for example, 136 frames per second. The time point within the cardiac cycle at which each image has been acquired is indicated by the solid ball below each image. As above for the low frame rate images, the strain is overlayed in color on the B-mode images.

Similarly, FIGS. 22(o)-22(ap) and 23(o)-23(ap) are greyscale images corresponding to FIGS. 22(a)-22(n) and 23(a)-23(n), respectively. In the greyscale images of FIGS. 22(o)-22(ap) and 23(o)-23(ap), the strain has been separated from the B-mode images for ease of viewing. Thus each image in FIGS. 22(a)-22(n) and 23(a)-23(n) corresponds to two images in FIGS. 22(o)-22(ap) and 23(o)-23(ap)—one for the B-mode image, the other for the strain.

Exemplary Source Code and Exemplary Systems—Conventional Ultrasound Machines

As referenced above, the computer program listing appendix includes a set of exemplary source code files implementing an exemplary embodiment of the present invention. The exemplary code is written in Matlab™, and implements various sub-processes depicted in FIG. 3, and was used to generate the exemplary images depicted in FIGS. 4(a)-14(b), as described above. The code was implemented on a conventional ultrasound machine, and can be implemented or adapted to process signals obtained from most standard ultrasound machines using known techniques.

List of Filenames:

analyzeNrf.m

cutECG.m

findSectorOverlap.m

matchbestECG.m

data2rgb.m

initOverlay.mn

overlayData.m

overlayimage.m

tiss2rgb.m

makeCardiacMovie.m

getPolTransformMap.m

appPolTransform.m

getmyparams.m

Additionally, the exemplary code provided in the computer program listing appendix can, for example, be adapted for use in a programmable ultrasound machine, such as for example, the Ultrasonix Sonix RP system. The Sonix RP system, for example, offers frame rate capabilities up to 700 fps as well as access to the beamformer. This higher frame rate can, for example, ensure higher strain quality, as has been seen in a preliminary in vivo human study performed by the inventors. In addition to the higher frame rate advantages of such a platform, access to the beamformer not only allows for the selection of optimal acoustic parameters, such as, for example, frequency, aperture and beamwidth, but it can also, for example, allow for further automation of the methods of the present invention. An exemplary implementation of the present invention on such a platform was performed by the inventors, as next described.

Examplary Implementation—Programmable Ultrasound Machine

High frame-rate ultrasound Radio-Frequency (RF) data acquisition is critical for myocardial elastography and imaging of the transient electromechanical wave propagation in cardiovascular tissues. To overcome the frame-rate limitations on routine ultrasound systems, the inventors developed an automated method for retrospective, multi-sector acquisition through synchronized electrocardiogram (ECG) gating on a clinical Ultrasonix RP system (Ultrasonix Medical Corp. Burnaby, Canada). A computer multithread technique was applied to acquire ECG and ultrasound RF signals simultaneously. The method achieved high spatial resolution (64-line beam density) and high temporal resolution (frame rate of 481 Hz) at a total imaging depth of 11 cm, 100% full view. A normal human heart left ventricle and a normal aorta were imaged using the same technique in vivo. Composite RF and B-scan full view frames were reconstructed by retrospectively combining all small-sector RF signals. The in-plane (lateral and axial) displacements of both long-axis and short-axis views of a healthy human left ventricle were calculated using an RF-based elastographic technique comprising 1D cross-correlation and recorrelation methods (windows size 6.9 mm, overlap 80%). A sequence of the electromechanical activation of the heart was observed through mechanical pulse waves propagating along septum (from base to apex) and posterior wall (from apex to base) during systole in human in vivo. Exemplary embodiments of this technique can, for example, expand the potential of echocardiography for quantitatively noninvasive diagnosis of cardiovascular diseases such as, for example, myocardial infarction, aneurism and early stage atherosclerosis. Heart diseases, such as ischemia and infarction, are a growing problem world wide. It is highly useful for the early diagnosis of such cardiac disease to noninvasively detect abnormal patterns of regional myocardial deformation caused by malfunction of the electromechanical conduction. Magnetic resonance (MR) cardiac tagging has been shown capable of quantifying the mechanical properties of the myocardium at high precision. However, the relatively low spatial resolution and the low temporal resolution limited the use of tagged magnetic resonance imaging (tMRI) for the detection of the transient mechanical vibrations that are constantly generated by the heart and the arteries. Therefore, echocardiography has been the predominant imaging modality in diagnostic cardiology owning to its real time, high temporal resolution capability. Tissue doppler imaging (TDI), strain rate imaging (SRI) and elastography imaging have been introduced to image the regional motion of the myocardium non-invasively. However, their major applications remain in the global motion of the heart over a complete cardiac cycle due to the current low frame rate.

In echocardiology, a high temporal resolution, typically <5 ms, is required to observe the detailed myocardium activities, such as, for example, the fast electrical conductive sequencing pattern for early detection of cardiac diseases. The electrical excitation, which induces the contraction and relaxation of the cardiac muscle, results in a strong electromechanical wave that propagates in the myocardium at a speed up to 5 m/s. Several methods had been developed to increase the ultrasound frame rate such as coded-excitation ultrasound imaging and parallel processing techniques. Most often these methods sacrificed field of view or ultrasound beam number to increase frame rate. This is not favorable in clinical study and is not optimal in general. ECG triggering or gating can be used to achieve high frame rate by reconstructing RF lines at different cardiac cycle especially for large field of view and high spatial resolution. The assumption of ECG triggering or gating lies in that the heart rate does not vary significantly, and that the myocardial function is effectively identical at every cardiac cycle. As was observed (and as shown in FIG. 26), ECG signals were very similar during systole for multiple cycles but could have up to a 10% length difference during diastole. Thus, all ECGs and corresponding RF frames taken for different sectors were interpolated to the same length to get the maximum similarity for each cardiac cycle. In contrast to conventional methods that transfer ECG signals to an arbitrary waveform generator and use the ECG R-wave as a trigger to control the transmitted pulses with synchronization implemented in hardware, a computer multithread technique for data acquisition was applied that significantly lowered the system cost without losing synchronization resolution.

High frequency, high resolution small animal ultrasound systems have become commercially available, such as, for example, the Vevo 770 system (VisualSonics Inc. Toronto, Ontario, Canada). However, there is still lack of high frame rate clinical systems for human cardiovascular study. This is largely because most commercial ultrasound systems are used effectively in clinical specialty areas where B-mode images are evaluated as the “gold standard.” Valuable frequency and speckle information carried by the RF echo signals is lost during conversion and compression, which occurs internally in the system. The Sonix RP system (Ultrasonix Medical Corporation, Burnaby, BC, Canada) is an open architecture system which can allow developers to easily control system parameters such as beam line density, sector size, and digitized RF signal acquisition etc. In the exemplary implementation, an Ultrasonix 500RP research platform was used to measure the ultrasound backscatter and attenuation coefficient. Using a Sonix RP ultrasound system, an elasticity imaging method with a frame rate (480 Hz) five times higher than traditional ultrasound machine (˜90 Hz) was obtained. 64 lines were kept for full sector view to reserve the high lateral resolution. The region of interest (ROI) was initially decreased to achieve the high frame rate. Then, an ECG gating technique was applied to utilize RF signals acquired during multiple cardiac cycles to retrospectively reconstruct small ROIs to a complete 100% full view cine-loop. Digitized RF and ECG signals were acquired through two computer threads running in parallel. In the exemplary implementation, local displacements were typically computed offline by applying a cross-correlation method to the pre compression ultrasonic radio frequency (RF) echo signals. Displacements were then estimated along the beam axis and displayed as an image referred to as an elastogram. The results obtained clearly showed electromechanical wave propagation in human heart during systole and a pulse wave propagating along a human aorta.

Data Acquisition

A clinical phased array transducer (Ultrasonix model # P4-2/20) operating at 3.3 MHz was used for human cardiac and vascular imaging. In a phased array transducer, more than one line can be acquired at the same time rather than line-by-line data acquisition by a signal element transducer to achieve high data consistency. For further development, if sector size is decreased to only one transmission line, the method could be reduced to single element scan imaging with an even higher frame rate. A separate ECG module (MCC Gesellschaft für in Medizin and Technik mbH & Co.KG.) was connected to the Sonix RP computer base running windows XP with RS232 serial interface. Two channels were recorded, from which three Einthoven and Goldberger leads were depicted. The signals were recorded digitally, processed and transmitted to the host via PC serial interface with a baud rate of 9600 bauds (1 start bit, 8 data bits, 1 stop bit, no parity). The maximum sampling frequency rate for this ECG module is 300 points per second.

Since the ultrasound RF and ECG data were two separate modules, synchronized data acquisition was critical for multi-sector combination method to achieve high frame rate through ECG gating technique. An automated custom designed program was developed in C++ based on the Ulterius SDK API to control, for example, both the Sonix RP system and the ECG module for synchronized ultrasound radio frequency (RF) and electrocardiogram (ECG) signal acquisition. Digitized ECG signals were stored in a computer using multi-thread technique which was in synchronization with ultrasound RF signal.

FIG. 24 provides exemplary process flow charts from which an exemplary C++ program that was created. The flow charts illustrate functionality for RF and ECG signal acquisition: a) Ultrasonix RF data acquisition. b) ECG module data acquisition. c) ECG and ECG time stamp buffer. d) RF frame and RF frame time stamp buffer.

When imaging began, raw ultrasound RF data for each frame was stored in the system cine buffer and retrieved by a custom designed “Callback” function. The “Callback” function had the capability of recording the occurrence time of each frame by a “::QueryPerformanceCounter( )” function. This time was used as a reference for ECG gating. The ultrasound RF signals were then stored line by line and then frame by frame in the computer memory sequentially. A separate computer thread running in parallel with the “Callback” function retrieved the digitized ECG signal points with its occurrence time from the serial port buffer. The total CPU usage was 70% on an Intel Pentium 4 2.99 GHz CPU based syctem, with 4 Gbyte memory. All RF and ECG signals were acquired and stored in real time. Both threads started to store data in physical memory simultaneously when the start-recording flag was set to TRUE(1), and stopped when the flag was set to FALSE(0). All RF and ECG signals for the various different sectors were saved in memory first and then written to the computer hard drive after the scanning. The total memory allocation for the data acquisition program was around 700 MB.

Composite Processing Through ECG R Wave Gating

For each sector, two sets of data, an ultrasound RF signal at a frame rate of 481 Hz and an ECG signal at 300 points per second, were recorded through two threads as described above. For every sector, the R-waves of the ECG signals were obtained through automatic peak detection. The corresponding time stamps of the R-wave peaks were used to search for the RF frame with the occurrence time most closely to the R wave occurrence time by the following method:

Min(ABS(Time_stamp_RF(i)−Time_stamp_ECG—R-wave)) i=1, 2, . . . , N

min(abs(Time_Stamp_RF(i)−Time_Stamp_{ECG—R-wavre})) i=1, 2, . . . , N,

where Time_Stamp_ECG—R-wave denotes the occurrence time of one of the ECG R wave peaks per sector, Time_Stamp_RF denotes the RF frame occurrence time, N is the total number of RF frames acquired together with ECG, and min and abs are Matlab functions used to find the minimum and absolute values of an array.

RF frames between two consecutive ECG R-waves were extracted from the original file representing one cycle of cardiac motion. Ideally the ECG cycle corresponding to each sector can be of identical duration to those of other sectors. The ith (i=1, 2, . . . ) frames of each sector can then be recombined in sequence to get the ith frame of the 100% view as is illustrated in FIG. 25. However one challenge for any ECG gated/triggered retrospective high frame rate ultrasound B mode imaging is heart rate variability. As noted, the duration of an ECG can vary by up to 10% per cycle, and the number of frames for each sector varies accordingly. As shown in FIG. 26, ECG signals during systole have very little variation. Therefore, an accurate method to solve the ECG arrhythmic is to stretch the diastolic part of the ECGs and the corresponding RF frames to the same length to achieve high similarity. One useful measure for the duration of systole is approximately T_es=√ΔT·0.343 s, where ΔT is the length in seconds of the cardiac cycle. Thus the systolic part of ECG can remain unchanged and the diastolic part of the ECG was interpolated to the maximum length of all seven ECG signals. The corresponding 2D RF frames were also linearly interpolated to the maximum length of the all sector RF frame sequence for better recombination.

FIG. 25 illustrates an ECG-gated multi-sector combination technique for high frame rate, full-view ultrasound imaging. In the example of FIG. 25, a total of seven sectors at different angles were acquired in a continuous sequence during each experiment. ECG and RF frame data are cut according to the time stamp associated with each data point or frame for one cardiac cycle. Corresponding small sector frames were, for example, recombined to generate full view ultrasound images.

FIG. 26 illustrates irregular ECG interpolation. ECGs during systole, approximately T_es=√ΔT·0.343 s, remain unchanged where T_es is the duration of the systole and ΔT is the duration of the whole cardiac cycle. All seven ECGs after the slashed line were, for example, linearly interpolated to the maximum length of these signals. The corresponding RF frames associated with each ECG were also interpolated to the maximum length of these RF by linearly 2D interpolation.

Motion Estimation

The axial displacement was estimated off-line using the normalized cross-correlation. The RF window size was equal to 6.9 mm and the window overlap was equal to 80%, deemed enough to retain good axial resolution. In the displacement estimation, the parabolic interpolation was applied to the cross-correlation function in order to further improve the precision. The displacements were then estimated using pairs of consecutive RF frames. To reduce the noise amplification effect of the gradient operator in the strain calculation, a linear Savitzky-Golay differentiation filter with a length of seven samples (140 um) was used to estimate the axial strains from the displacements. The aforementioned displacements and strains were the instantaneous or incremental displacements and strains occurring between two consecutive frames. Using the incremental displacements over one cardiac cycle, the preset points in the LV wall could be tracked automatically. Therefore, the incremental displacements and strains corresponding to the preset points could be extracted. By accumulating these incremental displacements and strains, the cumulative displacement and strains were obtained and represented the total motion and deformation from the first frame (corresponding to the first R-wave of the ECG), respectively. The displacements were color-coded and superimposed onto the grayscale B-mode images using an overlay blending mode. In the displacement images, only the estimates in the region of interest (ROI) are shown for better interpretation. An ROI was first determined through a 40- to 50-point selection performed manually in the first frame of a B-mode cine-loop (reconstructed from the RF image sequence). The selected points coincided with the myocardial boundaries (i.e., endocardium and epicardium). Using the estimated displacement field, these points could then, for example, be tracked over the entire cardiac cycle, providing the updated ROIs corresponding to different phases. The cumulative strain curve in myocardial elastography may undergo a drift, i.e., the cumulative strain does not return to zero at the end of the cardiac cycle. Thus, the drift in the cumulative displacements and strains was corrected on the assumption that the drift increases linearly with time over the duration of a cardiac cycle. Elastographic methods were implemented in MATLAB 7.1 (MathWorks, Inc., Natick, Mass., USA). The total processing time for a full cardiac cycle in the exemplary implementation was 2 to 3 hours on a PC workstation (Pentium 4 CPU 2.80 GHz, 2 GB RAM).

In Vivo Experiments

An adult healthy female heart was scanned both long axis and short axis view and the aorta with a frame rate of 481 Hz per sector through the custom automated program. The scan was performed with regular clinical ultrasound B-mode scan procedure by an experienced medical sonographer. The system parameters were set at 11 cm acquisition depth and a total of 64 lines for full 100% view. Ultrasound probe frequency was set at 3.3 MHz. Seven sectors were scanned with each sector of a 2-sec scanning time. At this time period, one or two cardiac cycles were recorded since the volunteer's heart rate ranged from 60 to 80 cycles per second. A total of 21-sec was needed for the entire experiment including scanning and data saving. Because respiratory motion can affect the heart position, breath holding was required for better composite images quality during the sector scanning. All seven sectors data was stored in memory and saved to hard drive after the scanning was completed. The patient's heart recovers to the original condition as much as possible during each cardiac cycle and the operator's hand keep still is essential to reconstruct a smooth transition from sector to sector. It is noted that although the total scanning time for a 100% 90 degree B-mode view is minimized by automatically sweeping different sectors, a patient's heart rate variability, breathing and the hand-freed motion of the transducer probe can pose some issues for accurate combination.

Results

FIGS. 27( a)-27(b) depict a comparison of image quality before and after overlap processing of different sectors. FIG. 27(a) depicts an example of a seven-sector composite B-mode image of a subject's heart long axis view. A 20% sector was scanned with a sector angle from 36 degree to −36 degree and with a −12 degree increment on Ultrasonix RP system with custom programmed software. A total of seven sectors were acquired, each sector having 12 RF lines. Three lines were cut for each sector since overlap exists between consecutive sectors. The total line density of the full composite view was (12−3)*7=63. An equivalent B-mode image of 100% one sector view is shown on FIG. 27(b) but at lower frame rate (90 Hz). All B-mode images were reconstructed using the Hilbert transform on the corresponding RF signals through Matlab function. Both images were acquired from start-systole. The quality of the two images is nearly identical and there are no clear transitions from sector to sector which validate the correct RF data acquisition and successfully recombination from ECG gated method.

After successful combination of the seven sectors RF data through ECG gating, the motion of the tissue was estimated off-line using an established classical speckle-tracking method. This technique was based on detecting the small local displacement of the tissue that occurs between tow consecutive frames. With the current method, only axial displacements (along the axis of the transducer beam), which coincided with the radial displacement in a long-axis view, were estimated. In our algorithm, the time shifts in the backscattered signals were determined between the two consecutive frames through cross-correlation of small sliding windows over the entire RF-line. This technique allowed the detection of very small displacement on the order of 0.1 um or less (correlation window of 6.9 mm, overlapping 80%). Finally, the cine-loop of the axial displacement was generated at a frame rate 481 Hz for the entire in vivo human cardiac cycle. Obviously there are complex movements, torsional, sagittal, and horizontal movements, especially during cardiac contraction. However, as shown by the magnetic resonance tagging pattern, there are far fewer torsional, sagittal, and horizontal movements in the IVS compared with those in the right ventricular anterior wall, the left ventricle (LV) posterior wall, and the apex. Thus, our measurements have been applied to the IVS to eliminate complexity of the three-dimensional motions during the cardiac cycle. FIGS. 29(a)-29(l) depict a sequence of color-coded axial displacements overlaid onto gray-scale B-mode images at different occurrence times during systole on a human left ventricle. In the displacement images, positive displacements (in red) denoted motion towards transducer whereas negative displacement (in blue) motion away from the transducer. A wave, known as electromechanical wave in red pointed by white arrows, was clearly seen traveling from the apex to the base right before the whole left ventricle start to contract. Previous experimental studies suggested that a contraction wave travels longitudinally on the LV epicardial surface from the apex toward the base. (From a physiologic view point, it is reasonable that the base contracts after the apical part, as a reverse pattern would squeeze the blood in the direction of the apex away from the aortic valve).

A “well-organized” heterogeneity in electromechanical coupling is thus a characteristic feature and may be a prerequisite for normal performance of the cardiac muscle.

In short axis view of the subject, another clear wave was also found propagating counter clockwise from septum to posterior wall during diastole phase as shown in FIGS. 30(a)-30(l). The wave front is indicated by white arrows.

Discussion

The first frame RF data and the first ECG data point are start at the same time as describe in the previous session. which could result in a maximum latency of 3.3 ms between the two data sets. This latency is determined by the maximum time interval of ECG sampling rate and the RF frame rate. In the worst case the latency between ECG and rf frames is min(1/ECGframe rate, 1/rf frame rate) since the first point of ECG data and the first frame of rf data is forced to be aligned.

L=Max(1/r_ECG, 1/r_RF)˜3.3 ms

Where L—latency of first RF frame and first ECG data

rECG—ECG sampling rate, 300 points/second

rRF—RF frame rate, >360 Hz

The ECG R wave peak position is detected by a matlab program where the corresponding RF frame position is calculated by

P _rf =P _r×round(N_rf/N_ECG)

Where

P_rf—the position of rf frame corresponding to ECG R wave peak

P_r—the ECG R wave peak position

N_rf—total number of rf frames.

N_ECG—total number of ECG points.

round—the matlab function to get the closest integer of N_rf/N_ECG

From FIGS. 29(a)-29(l), it was seen that at the end of diasystole, the contraction activation sequence in the axial direction is from apex to base with a relatively slow speed. This contraction pattern could be beneficial for left ventricle function, because it will first initiate the acceleration of the blood in LV to move toward the aorta exit from the ventricle chamber. High temporal resolution contrast echocardiography which bubbles are tracked in time and space for creating trajectories of blood in 2D had confirmed this blood flow accelerates in the apex-to-base direction, paralleling the apex-to-base direction of electromechanical activation.

A clear understanding of this type of ventricle mechanics is very useful clinically. It is also expected that this type of contraction synchronization would be the first to fail for heart failure diseases. This may be due to the fact that in the heart systolic contraction is an electrically triggered active event whereas diastole is a passive relaxation process.

Thus, in the above-described example, a method with a frame rate five times higher than obtainable using a conventional ultrasound imaging system was obtained. An ECG gating technique was applied for multi-sector recombination of RF signal to generate a 100% full view of the B-mode images.

Exemplary Pseudocode for ECG and RF Data Acquisition from Sonix RP System (Ultrasonix Medical Inc. Canada)

The following is exemplary pseudocode used in the above-described exemplary implementation.

//main program
program starts
parameter initializing;
declare message map;
if (!connected_to_Sonix RP host)
connect_to_host;
endif
detect ECG connection;
create ECG acquisition thread;
while(!exit) wait for user message;
program ends
// on catch message RUN_SCAN
retrieve scan parameter from panel;
if(!connected_to_host) error;
if(!ECG_running) error;
set scan parameter;
clear memory;
set_data_to_acquire(RF);
if(thread_scan exit)
terminate(thread_scan);
else
create_thread(thread_scan);
endif
//pseudo code for thread_scan
calculate total number of sectors;
for(i=starting_angle;i<stoping angle; angle_increment)
set sector angle;
post message starting scan
wait 2 s
post message stop scan
save ECG data in memory;
save RF data in memory;
endif

In a preliminary frame rate study conducted by the inventors, three conclusions were drawn: (1) during systole, the minimum frame rate for reliable strain information is approximately 250 fps. This is because the correlation coefficient surpasses 0.9, when the SNR_e is high enough (above 10 dB) for best images. This agrees with what various researchers have reported for cardiac RF speckle tracking, i.e., that the optimal frame rate is within the range of 200-300. (2) during diastole (as the strain rates during fast filling can be up to 2-3 times higher than during systole in humans), the minimum frame rate is approximately 500 fps. The optimal frame rate is directly proportional to the strain and strain rate amplitudes to be estimated. For fast filling, the strain rate is 2-3 times higher, therefore, the optimal frame rate needs to be accordingly adjusted; and (3) the Ultrasonix system can provide sufficiently high correlation coefficients (ρ>0.985), both for systolic and diastolic estimates. This result thus indicates that high correlation in a human heart is possible and that the most reliable strains are obtained at and beyond 250 fps for systole, and 500 fps for diastole, respectively.

While this invention has been described with reference to one or more exemplary embodiments thereof, it is not to be limited thereto and the appended claims are intended to be construed to encompass not only the specific forms and variants of the invention shown, but to further encompass such as may be devised by those skilled in the art without departing from the true scope of the invention.

Claims

1. A method of imaging of an anatomical object, comprising: acquiring a series of images for each of a plurality of sectors over one or more periods of a periodic signal representing movement of at least a portion of the anatomical object, each of the sectors corresponding to a different portion of the anatomical object;substantially simultaneously acquiring a period of the periodic signal in response to each periodic movement of the anatomical object, each of said series of images for each sector being acquired in response to the periodic signal; andsynchronizing data for each of said plurality of sectors using the periodic signal.

2. The method of claim 1, further comprising combining said data from said plurality of sectors to generate a series of high frame rate composite images.

3. The method of claim 2, further comprising displaying at least one of said series of high frame rate composite images to a user.

4. The method of claim 1, wherein at least one of said images comprises a 2D image.

5. The method of claim 1, wherein at least one of said images comprises a 3D image.

6. The method of claim 2, further comprising estimating one or more displacements between consecutive time frames for one or more windows of said high frame rate composite image.

7. The method of claim 6, further comprising estimating a propagation of said one or more displacements in a waveform.

8. The method of claim 7, further comprising estimating a direction of said propagation of said one or more displacement waves.

9. The method of claim 8, further comprising estimating a velocity of said propagation of said one or more displacement waves.

10. The method of claim 7, wherein said one or more displacement waves comprise one or more electromechanical waves.

11. The method of claim 10, further comprising estimating a direction of propagation of said one or more electromechanical waves.

12. The method of claim 11, further comprising estimating a velocity of propagation of said one or more electromechanical waves.

13. The method of claim 7, wherein said one or more displacement waves comprise one or more mechanical waves.

14. The method of claim 13, further comprising estimating a direction of propagation of said one or more mechanical waves.

15. The method of claim 13, further comprising estimating a velocity of propagation of said one or more mechanical waves.

16. The method of claim 6, further comprising applying a noise removal algorithm to said estimated one or more displacements.

17. The method of claim 16, further comprising accumulating said estimated one or more displacements with time to track motion for an entire period of movement of the anatomical object.

18. The method of claim 17, further comprising deriving strains in at least a portion of the anatomical object from said accumulated displacements.

19. The method of claim 18, further comprising overlaying data representing at least one of said displacement and said strain onto said high frame rate composite images.

20. The method of claim 1, wherein said synchronizing comprises leaving a first portion of each period of said periodic signal unchanged and interpolating a second portion of each period of said periodic signal to a total maximum length of the periods of the periodic signal.

21. The method of claim 20, further comprising linearly interpolating corresponding images to a maximum length of an all sector image sequence.

22. A computer program product comprising a computer usable medium having computer readable program code means embodied therein, said computer readable program code means in said computer program product comprising means for causing a computer to: cause an imaging device to acquire a series of images for each of a plurality of sectors over one or more periods of a periodic signal representing movement of at least a portion of an anatomical object, each of the sectors corresponding to a different portion of the anatomical object;substantially simultaneously cause a signal acquisition device to acquire a period of the periodic signal in response to each periodic movement of the anatomical object, each of said series of images for each sector being acquired in response to said periodic signal; andsynchronize data for each of said plurality of sectors using said periodic signal.

23. The computer program product of claim 22, said computer readable program code means in said computer program product further comprising means for causing a computer to: combine said data from said plurality of sectors to generate a series of composite images; anddisplay at least one of said series of composite images to a user.

24. An imaging system, comprising: at least one computer;at least one imaging device in communication with said at least one computer; anda periodic signal acquisition device, in communication with said at least one computer and said at least one imaging device, wherein said at least one computer causes said at least one imaging device to acquire a series of images for each of a plurality of sectors over one or more periods of a periodic signal representing movement of at least a portion of an anatomical object, each of said plurality of sectors corresponding to a different portion of the anatomical object; andwherein said at least one computer causes said periodic signal acquisition device to substantially simultaneously acquire a period of the periodic signal in response to each periodic movement of the anatomical object, each of said series of images for each sector being acquired in response to said periodic signal; andwherein the computer: synchronizes data for each of said plurality of sectors using said periodic signal.

25. The system of claim 24, wherein said at least one computer further combines said data from said plurality of sectors to generate a series of composite images.

26. The system of claim 25, wherein said at least one computer further processes said series of composite images to calculate accumulated displacement as a function of time for at least a portion of said anatomical object.

27. The system of claim 26, wherein said calculation of accumulated displacement includes estimating one or more displacements between consecutive time frames for one or more windows of said series of composite images.

28. The system of claim 26, wherein said calculation of accumulated displacement further comprises applying a noise removal algorithm to said estimated one or more displacements.

29. The system of claim 26, wherein said at least one computer further derives strain or strain rate in at least a portion of the anatomical object from said accumulated displacement.

30. The system of claim 29, wherein said at least one computer further overlays data representing at least one of said displacement, said strain and said strain rate onto said series of composite images.