automated estimation of cardiac motion using mri
TRANSCRIPT
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
1/28
AUTOMATED ESTIMATION OF
CARDIAC MOTION USING MRI
Guide : Prof. J.B. Jeeva,
Division of Biomedical Engg.
Place of work : VIT University
By
Sudhakar K (09MBE014),
M.Tech., Biomedical Engg.,
VIT University
1
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
2/28
AIM/OBJECTIVE
To segment the left ventricle from the Cardiac
MRI
To extract the clinically relevant parameters
capable of determining normal and abnormal
heart
To analyze the left ventricle motion using Optical
flow technique
2
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
3/28
METHODOLOGY
Read the Cardiac MRImages
Segment the Left
Ventricle Using
Morphological Operation
Find different
parameters
Find the MotionDirection of Left
Ventricle Wall
Analyze the normal and
abnormal motion of the
Left Ventricle
3
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
4/28
SEGMENTATION
Convert the RGB Image toBinary Image
Apply the Closing Operation
Find the Connected Pixels
Apply the IMFILL operation
to fill the opening in the leftventricle
Find the Edge of the Left
Ventricle 4
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
5/28
SEGMENTATION RESULTS
Fig.1 Short axis view
of Cardiac MRI
Fig. 2 Binary Image
Fig. 3 After appliedClosing Operation
Fig. 4 SegmentedLeft Ventricle
5
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
6/28
ABNORMAL LEFT VENTRICLE
Fig. 5 Ischemia
Fig. 6 Marfans
Fig. 7 VentricularTachycardia
6
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
7/28
PARAMETERS
Area: Number of pixels inside the contour of the
left ventricle
Perimeter: Number of pixels on the contour of
the left ventricle
2D Cross correlation: Computes the
correlation between two matrix
Variance: Computes the variance between the
two matrix
7
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
8/28
AREA OF LV
8
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
9/28
PERIMETER OF LV
9
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
10/28
2D CROSS CORRELATION BETWEEN TWO
SUCCESSIVE FRAMES
10
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
11/28
VARIANCE BETWEEN TWO SUCCESSIVE
FRAMES
11
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
12/28
MOTION ESTIMATION
Left Ventricle wall motion is estimated by optical
flow technique
Optical flow is the distribution of apparent
velocities of movement of brightness pattern in
an image
Optical flow method is used to calculate the
motion between two image frames which are
taken at time t and t+t at every pixel point
The basis of this method is intensity conservation
between consecutive images
12
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
13/28
FLOW CHART TO DETERMINE THE OPTICAL
FLOWTake 2 consecutive
Images
Preprocessing
Compute initialvelocity
Compute L1 & L2
from Initial velocity
Ite
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
14/28
C
Built Main Matrix
Solve Equation
Obtain Velocity and
set initial
velocity=velocity
Ite=Ite+1
14
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
15/28
PRE PROCESSING
Setting an external contour to the images of 2
pixels width, by doubling the pixels of the
original contour
To avoid getting a wrong result of thederivatives at the border of the image, and
thus, propagating this wrong result over the
pixels in the neighbourhood
Obtain the derivatives in in all the axes
15
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
16/28
COMPUTE INITIAL VELOCITY
The initial velocities computed are given by the
following equation, which is used to solve the
optical flow equation
minv(I
xv
1+I
yv
2+I
t)2
Where Ix,Iy- intensity of image at time t.
Taking the derivatives with respect to v1 and v2 (Ixv1+Iyv2+It)Ix=0
(Ixv1+Iyv2+It)Iy=0
16
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
17/28
CONTD.,
Clearing the Variables v1 and v2 so obtain
the initial velocitiesv1=IxIx.IyIt-IxIt.IxIy/-(IxIx.IyIy-IxIy
2)
v2=IyIy.IxIt-IyIt.IyIx/-(IxIx.IyIy-IxIy2
)
17
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
18/28
OPTICAL FLOW ESTIMATION
oTo solve this equation, will get the optical
flow between two consecutive images
oWhere v1TL1 (z)v1 = 0
v2TL2(z)v2= 0
From this, L1 & L2can obtain
18
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
19/28
RESULTS
Optical flow ofLeft Ventricle
wall (Frames
b/w 7th & 8th)
Optical flow
of Left
Ventricle
wall
(Frames b/w
11th & 12th)
Optical flowof Left
Ventricle
wall (Frames
b/w 21st &
22nd)
19
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
20/28
ABNORMAL MOTION
IschemiaRVOT
Enlargement
Ventricular
Tachycardia
20
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
21/28
PIXEL TRACKING
21
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
22/28
PARAMETERS OF NORMAL HEART
Frame Area Perimeter Variance 2D Cross
Correlation
1 1.3993e+03 148.0833 0 0
2 1.3993e+03 148.0833 0 1
3 1.3993e+03 148.0833 0 14 1.3675e+03 147.5980 1.0107e-06 0.5804
5 1.3675e+03 147.5980 1.0107e-06 1
6 1.3675e+03 147.5980 0 1
7 1.3088e+03 145.7401 1.1862e-06 0.5330
8 1.3088e+03 145.7401 1.1862e-06 1
9 1.3088e+03 145.7401 0 1
10 1.2244e+03 135.4975 1.8725e-06 0.3937
11 1.2244e+03 135.4975 1.8725e-06 1
12 1.2244e+03 135.4975 0 1
22
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
23/28
PARAMETERS OF ABNORMAL HEART
Frames Area Perimeter Variance 2D Cross
Correlation
1 859.1250 129.9828 0 0
2 690.2500 122.0833 2.1682e-06 0.1173
3 591 103.2548 1.9581e-07 0.30764 502 85.9411 2.0955e-09 0.2273
5 435.6250 75.6985 4.5635e-08 0.3125
6 387.6250 71.6985 8.3819e-09 0.1622
7 371.8750 70.5269 7.1304e-08 0.3901
8 367.7500 71.1127 7.5437e-08 0.6663
9 385.6250 73.6985 6.3388e-08 0.4181
10 486.3750 84.8701 8.8534e-08 0.2082
11 592.5000 99.5980 2.3283e-10 0.3250
12 709.2500 114.0833 5.8208e-11 0.3996
23
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
24/28
GUI IMPLEMENTATION
24
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
25/28
CONCLUSION
In this work the left ventricle is segmented using
thresholding and morphological operations
Optical flow technique is implemented to
estimate the left ventricle motion
The GUI is designed to help the clinician to
assess and analyse the motion of the left
ventricle
25
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
26/28
REFERENCES
Carranza-Herrezuelo, N. Bajo, A. Sroubek, F. and Santamarta, C (2010). Motionestimation of tagged cardiac magnetic resonance images using variationaltechniques, Computerized Medical Imaging and Graphics vol 34 pp 514522
Horn, B. K. P. and B. G. Schunk (1981). Determining Optical-Flow, ArtificialIntelligence, vol 17(1-3), pp 185-203
Lynch, M. Ghita, O. and Whelan, P.F. (2006). Automated segmentation of the leftventricle cavity and myocardium in MRI data, Elsevier Computers in Biology andMedicine,vol 34(4), pp 389-407.
MathWorks-Matlab and Simulation for Technical computing,, Accessed on 10th Jan 2011
Petros A. Maragos, Ronald W. Schafer, Morphological Skeleton Representation andCoding of Binary Images In Proc. Of IEEE transactions on acoustics. speech, andsignal processing, vol. assp-34, no. 5, Oct 1986
Pujadas, S., Reddy, G.P., Weber, O., Lee, J.J., Higgins, C.B. (2004). MR imagingassessment of cardiac function, Journal of Magnetic Resonance Imagingpp:789799
Rafael C.Gonzalez and Richard E. Woods, (2008). Digital Image Processing, Pearsoneducation, pp 1-750.
Soroor behbahani, Keivan magholi (2007). Analysis of cardiac wall motion estimationmethods, IEEE Trans., on medical imaging, pp.1102-1107
26
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
27/28
PUBLICATIONS
Sudhakar K and J.B. Jeeva (2011). Automated
Estimation of Cardiac Motion from tagged MRI
using CNN, International Conference on
Systemics, Cybernetics and Informatics vol 2(2),
pp 149-150
Sudhakar K and J.B. Jeeva (2010). Cardiac
Motion Analysis: A review VIT Science
Engineering and Technology Conference
27
-
8/6/2019 Automated Estimation of Cardiac Motion Using Mri
28/28
28