Language selection

Search

Patent 2861126 Summary

Third-party information liability

Some of the information on this Web page has been provided by external sources. The Government of Canada is not responsible for the accuracy, reliability or currency of the information supplied by external sources. Users wishing to rely upon this information should consult directly with the source of the information. Content provided by external sources is not subject to official languages, privacy and accessibility requirements.

Claims and Abstract availability

Any discrepancies in the text and image of the Claims and Abstract are due to differing posting times. Text of the Claims and Abstract are posted:

  • At the time the application is open to public inspection;
  • At the time of issue of the patent (grant).
(12) Patent: (11) CA 2861126
(54) English Title: IMAGE RESTORATION SYSTEM AND METHOD
(54) French Title: SYSTEME ET PROCEDE DE RESTAURATION D'IMAGE
Status: Granted
Bibliographic Data
(51) International Patent Classification (IPC):
  • A61B 8/00 (2006.01)
  • G06T 5/10 (2006.01)
  • G06T 5/00 (2006.01)
(72) Inventors :
  • BENAMEUR, SAID (Canada)
  • LAVOIE, FREDERIC (Canada)
(73) Owners :
  • EIFFEL MEDTECH INC. (Canada)
(71) Applicants :
  • EIFFEL MEDTECH INC. (Canada)
(74) Agent: PRAXIS
(74) Associate agent:
(45) Issued: 2020-11-10
(86) PCT Filing Date: 2013-01-23
(87) Open to Public Inspection: 2013-08-01
Examination requested: 2018-01-11
Availability of licence: N/A
(25) Language of filing: English

Patent Cooperation Treaty (PCT): Yes
(86) PCT Filing Number: PCT/CA2013/000057
(87) International Publication Number: WO2013/110174
(85) National Entry: 2014-07-14

(30) Application Priority Data:
Application No. Country/Territory Date
61/632,340 United States of America 2012-01-23
2,765,244 Canada 2012-01-23

Abstracts

English Abstract

A system and a method for improving the quality of ultrasound images. The system comprises a processor being configured to subdivide the ultrasound image, determine a deconvolution factor for the ultrasound image and apply the deconvolution factor to the subdivided ultrasound image, resulting in a restored image.


French Abstract

L'invention concerne un système et un procédé permettant d'améliorer la qualité d'images ultrasonores. Le système comprend un processeur étant configuré pour subdiviser l'image ultrasonore, pour déterminer un facteur de déconvolution pour l'image ultrasonore et pour appliquer le facteur de déconvolution à l'image ultrasonore subdivisée, ce qui permet de fournir une image restaurée.

Claims

Note: Claims are shown in the official language in which they were submitted.


18

CLAIMS
What is claimed is:
1. A system for improving the quality of an imaging system image,
comprising:
an input/output interface configured to receive the imaging system
image;
a processor in communication with the input/output interface, the
processor being configured to:
a) subdivide the imaging system image;
b) apply discrete cosine transform (DCT) denoising using a hard
thresholding rule;
c) apply an iterative expectation-maximisation (EM) regression
model;
d) estimate a point spread function;
e) set the DCT deconvolution factor to the estimated point spread
function.
f) apply the DCT deconvolution factor to the subdivided image;
and
g) provide a restored image based on the deconvoluted
subdivided image.
2. The system for improving the quality of an imaging system image of
claim 1, further comprising:
a display in communication with the input/output interface, the
input/output interface being further configured to provide the restored image
to
the display.

19

3. The system for improving the quality of an imaging system image of
claim 1, further comprising:
an imaging system in communication with the input/output interface, the
imaging system providing the imaging system image to the input/output
interface.
4. The system for improving the quality of an imaging system image of
claim 3, wherein the imaging system is an ultrasound imaging system.
5. The system for improving the quality of an imaging system image of
claim 3, wherein the imaging system is selected from a group consisting of a
radioscopic imaging system, a radiographic imaging system and an echographic
imaging system.
6. The system for improving the quality of an imaging system image of
claim 1, wherein the imaging system image is an ultrasound image.
7. The system for improving the quality of an imaging system image of
claim 1, wherein the imaging system image is selected from a group consisting
of a radioscopic image, a radiographic image and an echographic image.
8. The system for improving the quality of an imaging system image of
claim 1, further comprising the step of applying a homomorphic filter to the
associated modulation transfer function of the point spread function of the
imaging system image prior to step b).

20
9. The system for improving the quality of an imaging system image of
claim 1, wherein step b) consist in applying, iteratively, frequential
filtering
based on the DCT of 8 × 8 sub-images resulting from the subdividing of
the
imaging system image at step a).
10. The system for improving the quality of an imaging system image of
claim 1, wherein in step c) the EM regression model is set using the following

parametric form of the modulation transfer function in the Fourier domain:
H(u,v)= .pi..sigma.x.sigma.y exp(-2.pi.2.sigma.2xu2){exp(-2.pi.2.sigma.2y(v-
.function.)2 ) + exp(-2.pi.2.sigma.2y(v + .function.o)2)}
and applying an expectation-maximisation-based clustering algorithm to
the regression model.
11. The system for improving the quality of an imaging system image of
claim 1, wherein step f) is performed using an unsupervised Bayesian
deconvolution approach.
12. A method for improving the quality of an imaging system image,
comprising the steps of:
a) subdividing the imaging system image;
b) applying discrete cosine transform (DCT) denoising using a hard
thresholding rule;
c) applying an iterative expectation-maximisation (EM) regression
model;
d) estimating the point spread function;
e) setting the DCT deconvolution factor to the estimated point spread
function.
f) applying the DCT deconvolution factor to the subdivided image; and
g) providing a restored image based on the deconvoluted subdivided
image.

21
13. The method for improving the quality of an imaging system image of
claim 12, wherein the imaging system image is an ultrasound image.
14. The method for improving the quality of an imaging system image of
claim 12, wherein the imaging system image is selected from a group consisting

of a radioscopic image, a radiographic image and an echographic image.
15. The method for improving the quality of an imaging system image of
claim 12, further comprising the step of applying a homomorphic filter to the
associated modulation transfer function of the point spread function of the
imaging system image prior to step b).
16. The method for improving the quality of an imaging system image of
claim 12, wherein step b) consist in applying, iteratively, frequential
filtering
based on the DCT of 8 × 8 sub-images resulting from the subdividing of
the
imaging system image at step a).
17. The method for improving the quality of an imaging system image of
claim 12, wherein in step c) the EM regression model is set using the
following
parametric form of the modulation transfer function in the Fourier domain:
H(u, v) = .pi..sigma.x.sigma.y exp(-2.pi.2.sigma.2x u2){exp(-
2.pi.2.sigma.y(v¨.function.)2)+exp(-2.pi.2.sigma.2y(v+.function.o)2)}
and applying an expectation-maximisation-based clustering algorithm to
the regression model.
18. The method for improving the quality of an imaging system image of
claim 12, wherein step f) is performed using an unsupervised Bayesian
deconvolution approach.

22
Image

Description

Note: Descriptions are shown in the official language in which they were submitted.


CA 02861126 2014-07-14
WO 2013/110174
PCT/CA2013/000057
1
IMAGE RESTORATION SYSTEM AND METHOD
TECHNICAL FIELD
[0001] The present disclosure relates to an image restoration
system and method. More specifically, the present disclosure relates to
an image restoration system and method for restoring images obtained
from an ultrasound imaging system.
BACKGROUND
[0002] Contrary to other medical imaging techniques (e.g., X-rays,
magnetic resonance imaging, and computerized tomography), ultrasound
imagery is currently considered to be a non-invasive, portable, non-
expensive and safe (for the patient and operator) visualization medical
tool for investigating biological tissues of a body. However, despite
considerable advances in the technology of ultrasound imaging
equipment over the last years, the primary limitation of this imaging
modality remains its poor image quality (i.e. low signal-to-noise ratio, low
resolution and contrast), and also the presence of artifacts due to the
speckle noise effect that drastically deteriorates image quality and
sometimes makes imperceptible clinically important details within these
images (such as contours of anatomical structures).
[0003] In order to improve the quality of such ultrasound images,
an image deconvolution / restoration procedure could be efficiently
applied and, to this end, given a Point Spread Function (PSF) estimate,
many deconvolution models exist [1]. The only requirement for such
deconvolution algorithms consists, as a prerequisite first stage, of an
estimation of the PSF of the underlying ultrasound imaging system. This
problem of estimating the PSF and restoring is called a blind
deconvolution process and an alternative approach to this above-
mentioned estimation and deconvolution (disjoint) procedures consists of

CA 02861126 2014-07-14
WO 2013/110174
PCT/CA2013/000057
2
the simultaneous (generally iterative) estimation of the undegraded
original image and the PSF (or its inverse) [2-5].
[0004] Amongst the first blind deconvolution strategy for which
estimating the PSF estimation and the restoration process are two
disjoint procedures, there is the PSF identification procedure based on
frequency domain [6] zeros or the homomorphic filtering method which
consists in low-pass filtering (also called liftering) in the complex cepstral

domain (the cepstrum being defined by the inverse Fourier transform of
the log of the spectrum). This low-pass filtering is commonly achieved
either with an ideal low-pass filter [7, 8] or by hard or a soft shrinkage
rule
in the wavelet domain [9]. It is also worth mentioning the estimation
approach by means of local polynomial approximation proposed by Adam
and Michailovich [10], which can be viewed as a modification of
homomorphic estimation by using wavelet bases instead of the Fourier
basis. Nevertheless, ideal low-pass filtering in the cepstral domain or by
other wavelet-based filtering procedures have several drawbacks.
[0005] First, they are highly supervised to adequately set the
cutoff
frequency parameter which is crucial and different for each ultrasound
image because of the spatial variability of the PSF (due to the presence
of different interrogated tissues between the transducer and the
anatomical structure to be imaged).
[0006] Second, these classical filtering methods are not robust
enough to give a good estimate of the PSF spectrum and often tend to
produce artifacts in this estimation mainly due to the ringing effect of such
ideal low pass filter in the Fourier domain or due to the blocky effect
inherent to the wavelet based filtering procedure.
[0007] Accordingly, there is a need for an image restoration
system and method that addresses the above-described shortcomings.

CA 02861126 2014-07-14
WO 2013/110174
PCT/CA2013/000057
3
SUMMARY
[0008] The present disclosure provides a system for improving the
quality of an imaging system image, comprising:
an input/output interface configured to receive the imaging system
image;
a processor in communication with the input/output interface, the
processor being configured to:
a) subdivide the imaging system image;
b) determine a deconvolution factor for the imaging system
image;
c) apply the deconvolution factor to the subdivided image;
and
d) provide a restored image based on the deconvoluted
subdivided image.
[0009] There is further provided a system for improving the quality
of an imaging system image as above wherein step b) includes the sub-
steps of:
i) applying a homomorphic filter to the
associated modulation transfer function of the point
spread function of the imaging system image;
ii) applying denoising using a hard thresholding
rule;
iii) applying an iterative expectation-maximisation
regression model;
iv) estimating the point spread function; and
v) setting the deconvolution factor to the
estimated point spread function.

CA 02861126 2014-07-14
WO 2013/110174
PCT/CA2013/000057
4
[0010] The present disclosure also provides a corresponding
method for improving the quality of an imaging system image as well as a
processor executable product stored on a data storage medium,
configured to cause the processor to perform operations corresponding
to the method for improving the quality of an imaging system image.
BRIEF DESCRIPTION OF THE FIGURES
[0011] Embodiments of the disclosure will be described by way of
examples only with reference to the accompanying drawing, in which:
[0012] FIG. 1 is a schematic representation of an image restoration
system in accordance with an illustrative embodiment of the present
disclosure;
[0013] FIG. 2 is a schematic representation of an image restoration
system in a remote usage configuration;
[0014] FIG. 3 is a flow diagram of an image restoration process in
accordance with an illustrative embodiment of the present disclosure;
[0015] FIGS. 4A and 4B are ultrasound images of a distal femur
showing the medial side, corona! plane (FIG. 4A) and the medial
posterior condyle, axial plane (FIG. 4B);
[0016] FIGS. 5A and 5B show the modulus of Fr(u,v) after
application of the discrete cosine transform (DCT)-based denoising step
to the images of FIG. 4A and FIG. 4B, respectively;
[0017] FIGS. 6A and 6B are surface plots of the point-spread
function (PSF) defining a two-component mixture of bivariate Gaussian
distributions for FIG. 5A with p 154.18 134.21 ; 51.82 94.88] and
a =([358.66 4.18 ; 4.18 151.00], [358.84 4.10 ; 4.10 149.45]), and
FIG. 5A with p =153.05 131.53 ; 52.94 97,401 and a =([368.94-5.48 ;
5.48 97.40],[368.95 -5.47 ; -5.47 96.45]);

CA 02861126 2014-07-14
WO 2013/110174
PCT/CA2013/000057
[0018] FIGS. 7A to 7D are estimated spectrums of the point-
spread function (PSF) corresponding to FIG. 4A (FIGS. 7A and 7C) and
FIG. 4B (FIGS. 7B and 7D); and
[0019] FIGS. 8A and 88 are deconvolved images corresponding
5 to FIG. 4A and FIG. 4B, respectively.
[0020] Similar references used in different Figures denote similar
components.
DETAILED DESCRIPTION
[0021] Generally stated, the non-limitative illustrative embodiment
of the present disclosure provides a system and a method for improving
the quality of images obtained from an imaging system, such as an
ultrasound imaging system, through the application of an image
restoration process in order to recover clinically important image details,
which are often masked due to resolution limitations.
[0022] In common ultrasound imaging systems, the spatial
resolution is severely limited due to the effects of both the finite aperture
and overall bandwidth of ultrasound transducers and the non-negligible
width of the transmitted ultrasound beams. This low spatial resolution
remains the major limiting factor in the clinical usefulness of medical
ultrasound images.
[0023] To this end, an estimation of the Point Spread Function
(PSF) of the imaging system is required. The image restoration process
is a novel, original, reliable, and fast Maximum Likelihood (ML) approach
for recovering the PSF of an ultrasound imaging system. This new PSF
estimation method is based on an additional constraint, namely that the
PSF to be estimated is of known parametric form. Under this constraint,
the parameter values of its associated Modulation Transfer Function
(MTF) are then efficiently estimated using a homomorphic filter, a
denoising step, and an expectation-maximization (EM) based clustering

CA 02861126 2014-07-14
WO 2013/110174
PCT/CA2013/000057
6
algorithm. Consequently, this amounts to estimating, in the low-pass-
filtered cepstral domain, a mixture of two identical Gaussian distributions
whose parameters are automatically estimated, in a Maximum Likelihood
sense, by an iterative expectation-maximization (EM) [11] based
clustering algorithm. Given this PSF estimate, a deconvolution algorithm
can then be efficiently used, in a subsequent stage, in order to improve
the spatial resolution of ultrasound images, to obtain an estimate of the
true tissue reflectivity function, which is then independent of the
properties of the imaging system.
[0024] Referring to FIG. 1, the image restoration system 10
includes a processor 12 with an associated memory 14 having stored
therein processor executable instructions 16 for configuring the processor
12 to perform various processes, namely image restoration process,
which process will be further described .below. The image restoration
system 10 further includes an input/output (I/O) interface 18 for
communication with an imaging system 20 and a display 30.
[0025] The image restoration system 10 obtains images, for
example ultrasound images, from the imaging system 20 and executes
the image restoration process 16 on the acquired images. The resulting
restored images are then displayed on the display 30 and may be saved
to the memory 14, to other data storage devices or medium 40, or
provided to a further system via the I/O interface 18.
[0026] Referring to FIG. 2, the image restoration system 10 may
be remotely connected to one or more imaging systems 20 and/or
remotely operated through a remote station 62 via a wide area network
(WAN) such as, for example, Ethernet (broadband, high-speed), wireless
WiFi, cable Internet, satellite connection, cellular or satellite network,
etc.
The remote station 62 may also have associated data storage devices or
medium 64 for locally storing restored images provided by the image
restoration system 10.

CA 02861126 2014-07-14
WO 2013/110174
PCT/CA2013/000057
7
[0027] Referring now to FIG. 3, there is shown a flow diagram of
an illustrative example of the image restoration process 100 executed by
the processor 12 (see FIG. 1). Steps of the process 100 are indicated by
blocks 102 to 110.
[0028] The process 100 starts at block 102 where an image, for
example an ultrasound image, is obtained from the imaging system 20
and, at block 104, subdivided.
[0029] Then, at block 106, a deconvolution factor is determined for
the image and, at block 108, the deconvolution factor is applied to the
subdivided image resulting in a restored image.
[0030] Finally, at block 110, the restored image is provided, for
example through the display 30 and/or stored in a data storage device or
medium 40.
[0031] The various steps of process 100 will be further detailed
below.
PSF ESTIMATION BY HOMOMORPHIC TRANSFORMATION
[0032] In ultrasound imaging, the PSF happens to exhibit spatial
dependency due, among other things, to the non-uniformity of focusing,
the dispersive attenuation and the heterogeneity of the different
interrogated tissues. Nevertheless, a relatively low spatial variability of
these phenomena makes it possible to divide the obtained acoustic
image into a predefined number of small enough (possibly overlapping)
images, for which the data within each such smaller image can be
considered to be quasi-stationary, with a different PSF. It is then
assumed that, the entire image can be easily recovered by combining all
the local results obtained in this manner,
[0033] Assuming space invariance and linearity, the resolution
capabilities of an ultrasound imaging system can be expressed in terms

CA 02861126 2014-07-14
WO 2013/110174 PCT/CA2013/000057
8
of the PSF, h(x,y), i.e. the image of a point reflector, by the following
classical linear model:
g(x, y)= f (x,)* h(x, y) n(x,y) Equation .1
[0034] where f(x,y) is the spatial
reflectance distribution of internal
organs of the human body to be imaged, g(x,y) is the degraded
ultrasound image of the object f(x,y), h(x,y) is the PSF function of the
imaging system 20, which counts for the finite aperture and bandwidth of
the transducer, n(x,y) describes the additive quantization and electronic
noise and finally * designates the 2D discrete linear convolution operator.
Assuming that the noise term n(x,y) is temporarily ignored for the sake of
simplicity, Equation I is more easily described in frequency domain as a
simple product and sum where the capital letters indicate the Fourier
transforms of the corresponding spatial functions:
G(u,v)= F(u,v)H(u,v) Equation 2
[0035] An homomorphic
transformation is simply the complex
logarithmic transformation of both side of Equation 2. The real (Re) and
the imaginary (Im) parts of the resultant relation are given
correspondingly by:
Re : log p(u, v)I = log IF(u, v)1 + log IH(u, v)I Equation 3
lm 4G(u, v) = 4F(u, v) + 4H(u, v) Equation 4
[0036] where the symbols 1.1 and 4
denote, respectively, the
amplitude and the phase of the complex functions. The basic idea for
cepstrum-based methods of estimating the PSF spectrum H(u,v) relies
on the fact that log IH(u,v)I is typically a much smoother function than log
IF(u,v)I and the same holds for the functions 4H(u,v) and 4F(u,v).
Consequently, in this context, the log-spectrum of the degraded
ultrasound image (amplitude and phase) is considered to be a noisy
version of the complex log-spectrum of the PSF to be estimated and in

CA 02861126 2014-07-14
WO 2013/110174
PCT/CA2013/000057
9
this setting, in which log IF(u,v)1 and 4,F(u,v) are considered to be
sources of noise to be rejected, the problem of recovering log 1H(u,v)1
and 4H(u,v) is thus essentially a denoising problem in the cepstral
domain.
Denoising Step
[0037] In order to ensure both an automatic procedure and also a
reliable denoising step allowing a good estimate of the PSF spectrum,
H(u,v), without (ringing or blocking) artifacts, a two-stage denoising
scheme is proposed; namely a discrete cosine transform (DCT)-based
denoising step using a hard thresholding rule followed by a EM-based
regression model. In addition, since the PSF model relies on an even
function in x and y, the phase spectrum is assumed to be null.
DOT-based Denoisino Step
[0038] Algorithmically [12], the DOT-based denoising procedure
consists in applying iteratively, until a maximal number of iterations is
reached or until convergence is achieved, frequential filtering based on
the DCT transform of each 8 x 8 sub-image extracted from the current
version of the image to be denoised (initially, this current image estimate
is the noisy image itself). For the filtering operation in the DCT domain,
the easily-implemented hard thresholding rule [13] is used, also
classically used in wavelet based denoising approaches, where e is a
threshold level and w is one of the coefficients obtained by the OCT
transform of the block (of size 8 x 8 pixels) extracted from the current
image to be denoised. In order to reduce blocky artifacts across block
boundaries, a standard approach is adopted where this transform is
made translation-invariant, by using the DCT of all (circularly) translated
version of each channel of the image (herein assumed to be toroidal) [14]
(this implies computing a set of 8 horizontal shifts and 8 vertical shifts
transformed images) which is then averaged at each step of this iterative

CA 02861126 2014-07-14
WO 2013/110174
PCT/CA2013/000057
denoising procedure. In order to speed up the procedure, an overlap of
three pixels is used for the sliding 8 x 8 window. This iterative denoising
procedure, illustrated in Procedure 1, is applied on the noisy version of
log I-1(u,v), i.e., log G(u,v) (amplitude and phase) and allows us to obtain
5 a first rough estimate of log I-1"(u,v) which will be refined in the next
step.
Procedure 1 ¨ DCT-Based Denoising
Let
P-7 be the input image to be denoised at iteration n
P-7 be the denoised estimated image at iteration n
10 E be the threshold
For all (8 horizontal and 8 vertical) shifts of P' do
For all 8 x 8 blocks extracted from Pa do
1. OCT transform
2. Threshold the obtained OCT coefficients w with
the hard thresholding rule
= 0/f lw Is 6, w otherwise
3. Inverse OCT of these threshold coefficients
Unshift the filtered image and store it
1J3E-Averaging of these 64 denoised images
EM-based Estimation Step
[0039] In order to refine the estimation given by the above-
mentioned denoising step, the estimation method now relies on an
additional constraint, namely that the PSF to be estimated has the
following parametric form:
- X2 yz
h(x, y)= exp ¨ cos(27rfoy)
\20- 2ay
Equation 5
[0040] which is the PSF model used in [15], i.e. asymmetric
(across the x-axis and y-axis) cosine modulated by a Gaussian envelope

CA 02861126 2014-07-14
WO 2013/110174
PCT/CA2013/000057
11
whose the Fourier spectrum, i.e. its MTF (in fact a band-pass filter),
namely H(u,v) can be written in the Fourier domain:
H(u, v) =Ira y exp(-22r2cr.u2){exp(-2z2o-; (v - 4)2) + exp(-27r2o-;(v fo)2)}
Equation 6
[0041] Under this constraint,
the regression model that gives, for
the set of amplitude values of IH (u,v)I, the best fit, in the least square
sense, of two equally weighted Gaussian distributions (with the
constraints that these two distributions are centered at u = 0 and
symmetric with respect to v) can now be considered. In that respect, this
latter regression model can be efficiently addressed by considering the
parameter statistical estimation problem of a (noisy) Gaussian distribution
mixture of two (equally weighted) Gaussian component in R2 by
considering Nf 2-dimensional vectors v = (u,v)t , v = {vi, 15 i5 Nf }, taking
their values in R2 and whose cardinality of each v is given by the
amplitude value H(u,v). Finally, it is assumed that v = v1,..., vNF is a
realization in, IR2, of V whose density takes the form of the following 2-
component mixture:
2
P( v) EPkFm:(vic,,Pk) Equation 7
1,1
[0042] in which, the 2
components PV/Ci(Wck,Wk) are, in the
present application application (see Equation 5) assumed to be two
equally weighted (p1 = p2 = 0.5) bi-variate Gaussian distributions with
mean vector pk and identical covariance matrix 1(9ik =(pk,E)), i.e.:
P. = 1EN exP (v (v )}
Equation 8
[0043] In this setting,
the identification of the parameters of the
PSF spectrum modulus H(u,v) amounts to estimate the parameters (LH

CA 02861126 2014-07-14
WO 2013/110174
PCT/CA2013/000057
12
and 4)2 with the constraints that these two distributions are centered at u
= 0 (p1 = (u = 0, v1) and p2 = (u = 0, v2)) and v1 and v2 symmetric with
respect to v = 0, i.e. of opposite signs. This 2-component Gaussian
mixture model is estimated thanks to a EM-based clustering algorithm
[11]' The initial parameters of this iterative procedure are given by the ML
estimation on the partition given by a simple K-means clustering
procedure. The constraint of identical covariance matrix and mean vector
centered at u = 0 are taken into account at the end of the procedure by
simply considering the average value of the two covariance matrices and
the average absolute value of v1 and v2.
DECONVOLUTION
[0044] In order to improve the spatial resolution of the ultrasound
images and to obtain an estimate of the true tissue reflectivity function,
the ultrasound system's point-spread function can now be deconvolved
out. In the present application, an unsupervised Bayesian deconvolution
approach [16] is being used (or a penalized likelihood framework)
exploiting a non-parametric adaptive prior distribution derived from the
recent image model proposed by Buades [17]. This prior distribution
expresses that acceptable deconvolved solutions are the images
exhibiting a high degree of redundancy. In this setting, the deconvolution
of ultrasound images leads to the following cost function to be optimized:
E(f)=1g¨h*fil+Plf--r[g](f)1 Equation 9
[00451 where the first term expresses the fidelity to the available
data g and the second encodes the expected property of the true
undegraded image and Y[g](f) designates the non-local means filter in
[17] applied on f. p, the regularization parameter controlling the
contribution of the two terms (which is crucial in the determination of the
overall quality of the final estimate), is estimated with the method
proposed in [16].

CA 02861126 2014-07-14
WO 2013/110174
PCT/CA2013/000057
13
EXAMPLE
[0046] The PSF estimation approach and deconvolution were
texted on ultrasound images of several bones acquired using a portable
B-mode ultrasound imaging system (Titan, SonoSite Inc., Bothell, WA,
USA). The echographic appearance of the various tissues ranges from
dark (low-echoic) to bright (high-echoic), depending on their acoustic
impedance. FIGS. 4A and 4B show the original ultrasound images of the
distal femur, more specifically the medial side, coronal plane (FIG. 4A)
and the medial posterior condyle, axial plane (FIG. 4B)
[0047] FIGS. 5A and 5B show the modulus of H"(u,v) after
application of the DCT-based denoising step to the images of FIG. 4A
and FIG. 4B, respectively. It can be seen that two different pass-band
filters, related to two different PSFs are visible on these images. It can
also be seen that there is no aliasing error and this first denoising step
allowing the obtainment of the expected shape of a band-pass filter (see
Equation 5) on which the learning step of the Gaussian mixture,
exploiting the EM procedure, will be achieved, The Gaussian mixture,
estimated from these two spectrum data by the EM algorithm (without the
additional constraint of symmetry) is shown in FIGS. 6A and 6B. Two
examples of PSF estimation with the present approach are presented in
FIGS. 7A to 7D. Finally, FIGS. 8A and 8B show examples of
deconvolution ultrasound images using the deconvolution scheme
presented herein.
[0048] More specifically, FIGS. 6A and 6B are surface plots of the
point-spread function (PSF) defining a two-component mixture of
bivariate Gaussian distributions for FIG. 5A with p =[54.18 134.21 ; 51.82
94.88] and a =([358.66 4.18 ; 4.18 151.00], [358.84 4.10; 4.10 149.45]),
and FIG. 5A with p -153.05 131.53; 52.94 97.40] and a =([368.94-5.48;
5.48 97.40], [368.95 -5.47; -5.47 96.45]);

14
[0049] FIGS. 7A to 7D are estimated spectrums of the point-
spread function (PSF) corresponding to FIG. 4A (FIGS. 7A and 7C) and
FIG. 4B (FIGS. 7B and 7D), and FIGS. 8A and 8B are deconvolved
images corresponding to FIG. 4A and FIG. 4B, respectively.
[0050] Using the above-describe image restoration system and
method, greater resolution improvement of the deconvolved ultrasound
images can be observed with substantially improved definition of the
outer contour of biological structures and can easily be used for
commercial ultrasound applications due to its spatial resolution
improvement or as a prerequisite stage for the segmentation and 3D
reconstruction of ultrasound images.
[0051] It should be noted that although reference has been made
to ultrasound images and ultrasound imaging systems throughout the
present disclosure, it is to be understood that the image restoration
system and method may be applied and/or adapted to other types of
images and imaging systems such as, for example, radioscopic,
radiographic and echographic images from radioscopic, radiographic and
echographic imaging systems, or any other such images and imaging
systems.
[0052] Although the present disclosure has been described with a
certain degree of particularity and by way of an illustrative embodiments
and examples thereof, it is to be understood that the present disclosure is
not limited to the features of the embodiments described and illustrated
herein, but includes all variations and modifications within the scope and
spirit of the disclosure.
REFERENCES
[0053] In the present disclosure, references are made to the
following reference documents.
CA 2861126 2019-10-18

CA 02861126 2014-07-14
WO 2013/110174 PCT/CA2013/000057
[0054] [1] Mignotte, M., Meunier, J., Soucy, J.-P., and Janicki.,
C., "Comparison of deconvolution techniques using a distribution mixture
parameter estimation: application in spect imagery.," Journal of Electronic
Imaging 1, 11-25 (January 2002).
5 [0055] [2] Ayers, G. and Dainty, J., "Iterative blind
deconvolution method and its application," Optics Letters 13, 547-549
(July 1988).
[0056] [3] Katsaggelos, A. and Lay, K., "Maximum likelihood
blur identification and image restoration using the expectation-
10 maximization algorithm," IEEE Trans. on Signal Processing 39,729
¨733(March1991).
[0057] [4] Kundur, D. and Hatzinakos, D., "Blind image
restoration via recursive filtering using deterministic constraints," in
[Proc.
International Conference on Acoustics, Speech, and Signal Processing],
15 4,547-549 (1996).
[0058] [5] Benameur,S., Mignotte, M., Soucy, J.-P., and
Meunier, J., "Image restoration using functional and anatomical
information fusion with application to spect-mri images," International
Journal of Biomedical Imaging 2009, 12 pages(October 2009).
[0059] [6] Cannon, M., "Blind deconvolution of spatially
invariant image blurs with phase," IEEE Transactions on Acoustics,
Speech and Signal Processing 24, 58-63 (February 1976).
[0060] [7] Abeyratne, U., Petropulu, A., and Reid, J., "Higher
order spectra based deconvolution of ultrasound images," IEEE

CA 02861126 2014-07-14
WO 2013/110174
PCT/CA2013/000057
16
Transactions on Ultrasonics, Ferroelectrics and Frequency Control 42,
1064 ¨1075(Nov.1995).
[0061] [8] Taxt, T.,
"Restoration of medical ultrasound images
using two-dimensional homomorphic deconvolution," IEEE Transactions
on Ultrasonics, Ferroelectrics and Frequency Control 42, 543 554
(J uly1995).
[0062] [9]
Michailovich, 0. and Adam, D., "A novel approach to
the 2-d blind deconvolution problem in medical ultrasound," IEEE Trans.
on Medical Imaging 24, 86 ¨104 (Jan.2005).
[0063] [10] Adam, D. and
Michailovich, 0., "Blind
deconvolution of ultrasound sequences using nonparametric local
polynomial estimates of the pulse," IEEE Transactions on Biomedical
Engineering 49, 118 ¨ 131 (Feb. 2002).
[0064] [11] Dempster,
A., Laird,N., and Rubin, D., "Maximum
likelihood from incomplete data via the EM algorithm," Royal Statistical
Society, 1-38 (1976).
[0065] [121 Mignotte,
M., "Fusion of regularization terms for
image restoration," Journal of Electronic Imaging 19, 333004¨ (July-
Septem ber 2010).
[0066] [13] Donoho, D. L. and
Johnstone, I, M., "Ideal spatial
adaptation by wavelet shrinkage," Biometrika 81, 425-455 (1994).
[0067] [14]Coifman, R.
and Donohu, D., 'Translation in variant
denoising," in [Wavelets and Statistics, Lecture Notes in Statistics],

CA 02861126 2014-07-14
WO 2013/110174
PCT/CA2013/000057
17
103,125-150, A. Antoniadis and G. Oppenheim, Eds. New York:
Springer-Verlag ( 1995).
[0068] [15] Kallel, F., Bertrand, M., and Meunier, J., 'Speckle
motion artifact under tissue rotation," IEEE Transactions on Ultrasonics,
Ferroelectrics and Frequency Control 41,105 ¨122 (Jan.1994).
[0069] [16] Mignotte, M., "A non-local regularization strategy
for image deconvolution," Journal Pattern Recognition Letters 29(16),
2206-2212 (2008).
[0070] [17] Buades, A., Coll, B., and Morel, J.M,, "A review of
image denoising algorithms, with a new one," Multiscale Modeling and
Simulation (SIAM Interdisciplinary Journal) 4(2), 490-530 (2005).

Representative Drawing
A single figure which represents the drawing illustrating the invention.
Administrative Status

For a clearer understanding of the status of the application/patent presented on this page, the site Disclaimer , as well as the definitions for Patent , Administrative Status , Maintenance Fee  and Payment History  should be consulted.

Administrative Status

Title Date
Forecasted Issue Date 2020-11-10
(86) PCT Filing Date 2013-01-23
(87) PCT Publication Date 2013-08-01
(85) National Entry 2014-07-14
Examination Requested 2018-01-11
(45) Issued 2020-11-10

Abandonment History

There is no abandonment history.

Maintenance Fee

Last Payment of $263.14 was received on 2023-12-19


 Upcoming maintenance fee amounts

Description Date Amount
Next Payment if small entity fee 2025-01-23 $125.00
Next Payment if standard fee 2025-01-23 $347.00

Note : If the full payment has not been received on or before the date indicated, a further fee may be required which may be one of the following

  • the reinstatement fee;
  • the late payment fee; or
  • additional fee to reverse deemed expiry.

Patent fees are adjusted on the 1st of January every year. The amounts above are the current amounts if received by December 31 of the current year.
Please refer to the CIPO Patent Fees web page to see all current fee amounts.

Payment History

Fee Type Anniversary Year Due Date Amount Paid Paid Date
Application Fee $400.00 2014-07-14
Maintenance Fee - Application - New Act 2 2015-01-23 $100.00 2015-01-20
Maintenance Fee - Application - New Act 3 2016-01-25 $100.00 2016-01-19
Maintenance Fee - Application - New Act 4 2017-01-23 $100.00 2017-01-20
Request for Examination $200.00 2018-01-11
Maintenance Fee - Application - New Act 5 2018-01-23 $200.00 2018-01-11
Maintenance Fee - Application - New Act 6 2019-01-23 $200.00 2019-01-22
Maintenance Fee - Application - New Act 7 2020-01-23 $200.00 2020-01-22
Final Fee 2020-09-04 $300.00 2020-09-03
Maintenance Fee - Patent - New Act 8 2021-01-25 $204.00 2021-01-22
Maintenance Fee - Patent - New Act 9 2022-01-24 $203.59 2022-01-21
Maintenance Fee - Patent - New Act 10 2023-01-23 $263.14 2023-01-20
Maintenance Fee - Patent - New Act 11 2024-01-23 $263.14 2023-12-19
Owners on Record

Note: Records showing the ownership history in alphabetical order.

Current Owners on Record
EIFFEL MEDTECH INC.
Past Owners on Record
None
Past Owners that do not appear in the "Owners on Record" listing will appear in other documentation within the application.
Documents

To view selected files, please enter reCAPTCHA code :



To view images, click a link in the Document Description column. To download the documents, select one or more checkboxes in the first column and then click the "Download Selected in PDF format (Zip Archive)" or the "Download Selected as Single PDF" button.

List of published and non-published patent-specific documents on the CPD .

If you have any difficulty accessing content, you can call the Client Service Centre at 1-866-997-1936 or send them an e-mail at CIPO Client Service Centre.


Document
Description 
Date
(yyyy-mm-dd) 
Number of pages   Size of Image (KB) 
Maintenance Fee Payment 2020-01-22 1 33
Final Fee 2020-09-03 1 56
Representative Drawing 2020-10-14 1 4
Cover Page 2020-10-14 1 30
Maintenance Fee Payment 2021-01-22 1 33
Maintenance Fee Payment 2022-01-21 1 33
Maintenance Fee Payment 2023-01-20 1 33
Abstract 2014-07-14 2 62
Claims 2014-07-14 5 131
Drawings 2014-07-14 6 70
Description 2014-07-14 17 623
Representative Drawing 2014-09-08 1 5
Cover Page 2014-09-19 1 32
Request for Examination 2018-01-11 1 40
Maintenance Fee Payment 2018-01-11 1 40
Office Letter 2018-03-14 1 25
Examiner Requisition 2018-11-07 4 220
Maintenance Fee Payment 2019-01-22 1 33
Claims 2019-05-07 5 111
Amendment 2019-05-07 16 447
Interview Record Registered (Action) 2019-10-18 1 13
Amendment 2019-10-18 7 169
Description 2019-10-18 17 634
Claims 2019-10-18 5 110
PCT 2014-07-14 2 70
Assignment 2014-07-14 2 83
Fees 2015-01-20 1 25
Fees 2016-01-19 1 24
Maintenance Fee Payment 2017-01-20 1 27