Transcript SuYi 6.28Mbytes - The International Conference on

Biomedical Modelling with
Clinical Applications
Su Yi | 苏易
From an engineering perspective…
Designcentric
From a clinical perspective…
Patientcentric
An example in dealing with heart failure
Epidemic of chronic heart failure
among survivors of heart attack
What happens after an heart attack?
LV remodelling: the process in
which the heart alters its size,
shape & configuration after a
heart attack
Serial assessment & quantitation
of LV remodelling facilitates
– surveillance of heart failure
progression
– tailoring of appropriate treatment
& monitoring of efficacy
– reduce cost
How is LV remodelling being assessed?
Ejection Fraction (EF), i.e. change in
LV volume
Qualitative or semi-quantitative
descriptors of shape, e.g. sphericity
index (SI) and conicity index (CI)
Dimension, e.g. change in LV
diameter, wall thickness – imagebased
Twisting, e.g. tagged-MRI, speckle
tracking echocardiography – imagebased
Ventricular wall stiffness (σ/ε) –
FEM-based
How is LV remodelling being assessed?
Cardiac magnetic resonance (CMR)
imaging
– LV structure and function, e.g. LV
dimensions, LV volumes, etc
– Infarct size and extent – late
gadolinium enhancement (LGE)
Fails to exploit full range of
quantitative multi-dimensional MRI
data
LGE confirms LV apical infarct
Our idea…
2D images
3D model
4D spatial-temporal model
To develop new indices to quantitate LV remodelling using a
computational geometry approach
– Extract 3D/4D information
– Provide localised patient-specific details
Apply rigorous engineering and physiological principles to
derive cardiac indices which are robust and scientifically valid
Our approach…
What kind of clinical inference can we
make from the model?
Approximating local shape of LV
The aim of the surface-fitting process is to compute
an extended quadric of the form
z  ax 2  bxy  cy 2  dx  ey
which approximates the local geometry in the
vicinity of a point p on a mesh.
The local curvedness (or RMS curvature) is then
given by
C
 12   2 2
2
1 2 B 2  A 2 (4ac  b 2 )

A
A3
2
2
where A  d  e  1
B  a  ae 2  c  cd 2  bde
Extracting physical properties
Local 3D radius (R)
1
C
Local wall thickness (T) equivalent to solving a raytriangle intersection of a ray
with the epicardial surface
 d 
1
 e
nˆ 
 
d 2  e2  1  1 
 
R
Local wall stress ()
  0.133  SP 
R
T (2 
T
)
R
n̂
Is 3D method better than 2D image-based
methods?
10 normal subjects
Why is a 3D method better than 2D imagebased methods?
t1
r1
t2
r2
Regional curvedness
10 control; 10 diseased
Regional Wall Stress and Thickening
High wall stress,
especially at apex
Very little wall thickening
despite high wall stress
Validation against LGE
late gadolinium enhancement
representing myocardial scaring/fibrosis
Surgical Ventricular Restoration
40 patients; pre- and post-SVR
Before SVR
4 months after SVR
Comparison to other methods
Existing
Method
Imaging
modality
Type of indices
Our Method
Complexity
Semiqualitative
methods
Not restricted
Idealised local
and global
 Non-idealised, subjectspecific
Low
EF
Not restricted
Global
 Localized, regional,
global
Low
Thickening
Not restricted
Localized 2D
 Localized 3D, more
accurate
Low
Twisting
Tagged MRI,
Speckle
tracking
Localized
 (need to solve mesh
correspondence problem)
High
Strain (/ )
Not restricted
Localized

  (need to solve mesh
correspondence problem)
High
What are the implications?
Potential to replace delayed contrast
hyperenhancement MRI
Reduce scanning time/cost by at least 25%
Avoid the need to inject Gadolinium which might
result in complication in some patients
Reduce patient trauma
What are the current limitations?
Current approach does not exploit full 4D information; finite
difference between end-diastole and end-systole phase
Partitioning of endocardial surface according to medical
nomenclature assumes rigid rotation and linear vertical
compression; not realistic
Assumption of uniform pressure loading in LV chamber
What’s needed to bring this to the next level?
Clinical trial & testbedding
– 20 normal
– 30 diseased over time (at
least 1 year)
– age-matched
– gender-matched
Acceptance by clinicians
Medical Journals
–
–
Zhong L, Su Y, Yeo SY, Tan RS, Ghista DN, Kassab G. Three-dimensional curvedness and wall
stress assessment in dilated cardiomyopathy using magnetic resonance imaging. Am J Physiol
Heart Circ Physiol, 296: H573-H584, 2009.
Yeo SY, Zhong L, Su Y, Tan RS, Ghista DN. A curvature-based approach for left ventricular shape
analysis from cardiac magnetic resonance imaging. Med Biol Eng Comput, 47(3): 313-322, 2009.
Invited Book Chapter
–
Zhong L, Tan RS, Su Y, Yeo SY, Ghista DN, Kassab G. Noninvasive Assessment of Left
Ventricular Remodeling: Geometry, Wall Stress and Function, in Computational Cardiovascular
Mechanics: Modeling and Applications in Heart Failure, Julius Guccione and Mark Radcliff, Ed.
Medical Conferences
–
–
–
–
L. Zhong, Y. Su, S. Y. Yeo, D. Ghista, R. S. Tan. Three-dimensional left ventricular regional shape
and wall stress alterations after surgical ventricular restoration, accepted for oral presentation at
the 11th World Congress on Medical Physics and Biomedical Engineering, September 7-12, 2009
in Munich, Germany.
Yeo SY, Zhong L, Su Y, Tan RS, Ghista DN. Analysis of left ventricular surface deformation during
isovolumic contraction. Conf Proc IEEE Eng Biol Soc 2007;1:787-790. (EI, PubMed)
Zhong L, Yeo SY, Su Y, Le TT, Tan RS, Ghista DN. Regional assessment of left ventricular
surface shape from magnetic resonance imaging. Conf Proc IEEE Eng Biol Soc 2007;1:884-887.
(EI, PubMed)
Yeo SY, Tan RS, Chai GB, Ghista DN. Variation of left ventricular surface shape during the
cardiac cycle. The 3rd IASTED International Conference on Biomechanics, BioMech 2005.
Benidorm, 7-9 Sep 2005.
Some concluding thoughts…
Near confluence point of computational and
clinical practitioners
Need appropriate structure to facilitate
communications
Novelty vs Impact
Ease-of-use is key
Thank You