Evaluation Optical-Flow based Methods for Estimation of Wall Motions
1-Soroor Behbahani and 2- Keivan Magholi
Biomedical Engineering Department, Science and Research Branch, Islamic Azad University,
Tehran, Iran.
1-Member of Young Researcher Club of Biomedical Engineering Department, Science and Research
Branch, Islamic Azad University, Tehran, Iran.
soroor_behbahani@yahoo.com
Abstract
Cardiac disease remains a major killer in the world.
Improving the management of cardiovascular disease is
one of the greatest challenges facing healthcare. It is
valuable to study the electro-mechanics of the heart as
any asynchrony of electrical activation and disorder of the
mechanical properties can lead to abnormalities in heart
function. Accurate quantitative of heart motion and
deformation are of importance for evaluating normal and
abnormal cardiac physiology and mechanics. However,
the complexity of LV motion, the absence of internal
landmarks in the myocardium and the influence of the
right ventricle (RV) on the LV imply that the true motion
trajectories of tissue elements are difficult to obtain from
image analysis. In this paper we decided to compare two
methods of heart motion estimation and consider the
advantages or disadvantages of each method.
Keywords: motion estimation, myocardial motion, optical
flow, velocity field, medical imaging
1. Introduction
Estimation techniques of heart wall motion could
be categorized into invasive and non-invasive. Invasive
techniques track markers physically implanted on the
surface of the ventricle wall [9]. Movement of the
markers is tracked in sequence of cardiac images and
estimation of LV motion is possible. Such techniques are
not appropriate for wide application because of the need
for surgical intervention Non-invasive techniques
overcome this problem. There are three different groups
of approaches. The first group of approaches uses
magnetic resonance tagging technique where
magnetization of the tissue is altered such that grid of
intersecting planes is produced [11][13]. Points of
intersection are easily tracked as heart tissue moves, but
motion of other points is not estimated. The second group
of approaches analyzes shape of previously segmented
cardiac wall extracting motion information from the
changes in the shape [6][7].
Such approaches are usually limited to several
characteristic points extracted from the LV boundary. In
the third group of approaches are optical flow techniques
[12] which detect changes in brightness intensity of every
pixel in an image followed by estimation of movement
from detected data. The optical flow approach shows
some very good results in estimation of simple
movements but for complex motion observed in cardiac
images the algorithm need to be improved with additional
constraints. Other approach uses optical flow results as
input for further motion analysis [4]. Optical flow
algorithms provide estimation of the motion fields. In
general, optical flow information is not the same as the
motion field. The motion field is represented by the field
of vectors that show the displacement of points in the
optical field relative to the observer. Optical flow shows a
velocity field of pixels in the image. There are a number
of ways to compute the optical flow. The basic
classification of methods is of gradient-based ([16]),
correlation-based and energy-based methods ([8]).In this
paper we decided to compare the optical flow algorithm
based on and energy-based a point-constrained methods.
2. Methods
2.1. Energy-Based Method
This method is based on minimization of the
velocity field energy function that is computed for two
consecutive image frames. The minimal energy state
corresponds to the optimal optical flow estimation. The
energy function minimization is based on the steepest
descent algorithm. The energy function E used in this
algorithm consists of two terms, E1 and E2 are:
E = E1 + E2 (1)
Second International Multisymposium on Computer and Computational Sciences
0-7695-3039-7/07 $25.00 © 2007 IEEE
DOI 10.1109/IMSCCS.2007.78
164