3. “All models are wrong”Denis Noble.
“Nobody believes on the results of an
experimentalist, but him, and everybody
believes on the results of a modeller, but
himself”
Take home message: … need to define the
problem and hypothesis as clear as possible!!
6. Cardiac remodelling
Development
Disease
State of art: coarse metrics
Length, diameter, volume…
Opportunities
Myriad of shape patterns
Tons of data
7. Computational statistical atlas of anatomy [1]
Clinicians will adopt novel shape coordinates
in this parametric space
[1] A. Young, A. Frangi. “Computational cardiac atlases: from patient to population and
back.” Exp. Physiol. (2009)
8. [2] P. Lamata, S. Niederer, et al., “An accurate, fast and robust method to generate
patient-specific cubic Hermite meshes,” Med. image Anal. (2011).
[3] P. Lamata, M. Sinclair, et al. “An automatic service for the personalization of
ventricular cardiac meshes.” J R Soc Interface (2014)
Model: ellipsoid
Meshing [2,3]
Reduce noise and artifacts
Smooth C1 representation
Statistics: PCA
Web-service
http://amdb.isd.kcl.ac.uk/
9. Give me your short axis stack, and I’ll tell you
if you had a premature birth [4].
[4] A. Lewandovski, D. Augustine et al. “Preterm heart in adult life: cardiovascular
magnetic resonance reveals distinct differences in left ventricular mass, geometry, and
function.” Circulation (2013)
10. Ventricle grow differently depending on
surgical choice in HLHS [5].
[5] J. Wong, P. Lamata et al. “Right ventricular morphology and function following
stage I palliation with a modified Blalock-Taussig shunt versus a right ventricle-to-
pulmonary artery conduit” Circ. Imaging (in review)
11. [5] J. Wong, P. Lamata et al. “Right ventricular morphology and function following
stage I palliation with a modified Blalock-Taussig shunt versus a right ventricle-to-
pulmonary artery conduit” Circ. Imaging (in review)
12.
13. HF with Normal Ejection Fraction
Evidence of abnormal filling caused by
stiffer myocardium, delayed relaxation,
impaired atrio-ventricular conduit function.
Diagnostic surrogates [6]:
• Lab: natruiretic peptides
• Echo: ratio early/late filling
• Catheters: LV pressure
Stratification: on-going challenge [6]
[6] Maeder and Kaye, “Heart
Failure With Normal Left
Ventricular Ejection Fraction,” J.
Am. Coll. Cardiol. 2009
14. State of art (catheter): exponential fitting
Coupling between relaxation and stiffness
P
V
Passive elastic
Active fibre relaxation
Total LV pressure
15. Myocardial properties (relaxation/stiffness)
Input: deformation and pressure
Method: Model personalization
Output: Decouple relaxation / stiffness
16. [7] J. Xi, P. Lamata, et. al, “The estimation of patient-specific cardiac diastolic functions
from clinical measurements,” Med. image Anal., 17:133-146 (2013).
6 unknowns
4 data points
Additional constraints [7]
End diastole: null active tension
Positive, and monotonically decaying active tension
Criterion to choose reference configuration
17. [7] J. Xi, P. Lamata, et. al, “The estimation of patient-specific cardiac diastolic functions
from clinical measurements,” Med. image Anal., 17:133-146 (2013).
Criterion to choose reference configuration
18. Stiffness = f(deform., pressure)
LV filling pressure: only catheter
Two aims [8]:
Hypothesis: P = f(V)
Characterise impact of pressure
offset errors
[8] J. Xi, W. Shi, et. al, “Understanding the need of ventricular pressure for the estimation
of diastolic biomarkers,” Biomech. Model. Mechanobiol. (2013)
19. Literature surrogate
Able to differentiate
stiffness
Stiffness = f(ejection
fraction)
Unable to different.
active tension
[8] J. Xi, W. Shi, et. al, “Understanding the need of ventricular pressure for the estimation
of diastolic biomarkers,” Biomech. Model. Mechanobiol. (2013)
20. Able to recover from pressure offset errors
Need temporal resolution!
No pressure offset With pressure offset
[8] J. Xi, W. Shi, et. al, “Understanding the need of ventricular pressure for the estimation
of diastolic biomarkers,” Biomech. Model. Mechanobiol. (2013)
26. No need of
boundary conditions
Arbitrary domains
Includes viscous
effects
[8] S. Krittian, P. Lamata et al. “A FEM approach to the direct computation of relative
cardiovascular pressure from time-resolved MR velocity data.” Med. Im. Analysis (2012)
27. Mass and momentum conservation:
Viscous forces
Convective acceleration
(in-space)
Transient acceleration
(in-time)
Inertial forces
t=1
t=0
28. [9] P. Lamata, A. Pitcher et al. “Aortic relative pressure components derived from 4D flow
cardiovascular magnetic resonance”. MRM (2014)
29. Transient: pump action
and compliance
Convective: vessel
geometry
Viscous: inefficiencies
due to friction
[9] P. Lamata, A. Pitcher et al. “Aortic relative pressure components derived from 4D flow
cardiovascular magnetic resonance”. MRM (2014)
30.
31. Images are drivers of
modelling progress [10]
Complexity vs. clinical
adoption
Robustness!!
[10] P. Lamata, R. Casero et al, “Images as drivers of progress in cardiac
computational modelling”, Prog Biophys Mol Biol (2014)
32. Meshes of high quality [11]
[11] P. Lamata, I. Roy et al. “Quality metrics for high order meshes: analysis of the
mechanical simulation of the heart beat.” IEEE Trans Med Imag (2013)
33. More stable simulations: guide the optimizer
to enforce non-compressibility
[12] S. Land, S. Niederer et al. “Improving the stability of cardiac mechanical
simulations” IEEE Trans Biom Eng (accepted)
35. Oxford / KCL
Nic Smith
Steve Niederer
David Nordsletten
Sander Land
[Jiahe Xi]
[Sebastian Krittian]
[Ishani Roi]
Imperial
Daniel Rueckert
Wenzhe Shi
Clinicians
Reza Razavi (KCL)
Aldo Rinaldi (KCL)
Paul Leeson (OXF)
Adam Lewandovski (OXF)
Stefan Neubauer (OXF)
Alex Pitcher (OXF)
Editor's Notes
Thank you for the introduction, and for the opportunity to present my ideas and goals.
The problem that I will address is the management of HF, a major health issue in the UK, which brings annual costs of £0.75 billion to our health system. The lifetime risk of developing HF is one in five, it is expected that 3 of us will develop some form of this disease.
HF is the clinical condition in which the heart is not able to pump enough blood to meet the body demands. Two actions govern the mechanical pump function of the heart, ejection and filling. The scope of my work focuses on the second, which relates to the condition of HFNEF, that affects half of the population with HF (the other half have systolic HF). HFNEF patients have abnormal filling, caused by a stiffer myocardium, a delayed relaxation, or an impaired atrio-ventricular counduit function.
Current diagnostic clinical guidelines for HFNEG use these surrogates to characterise an impaired filling of the ventricles: 1,2,3
The problem is that the characterization and stratification of patients is an on-going challenge. One of the fundamental reasons for it is that current diagnostic metrics are only surrogates of the mechanisms that impair ventricular filling. In my work, I plan to bridges this gap, estimating the fundamental mechanical properties that govern diastolic filling.
The huge potential of the combination of models and images (observations)
Therefore, my central hypothesis is that these new mechanical parameters will improve the characterization and stratification of patients, and therefore the management of HF.
And therefore my objective is to develop a robust and clinically applicable methodology to characterise these mechanical parameters.
I will address this ambitious goal combining three of my most recent contributions in the field of research overlapping medical imaging and mathematical modelling.
First, the technology to personalise mechanical computational meshes to the anatomy of the patient, captured from medical images. A proof of the accuracy of this process, and also a significant contribution since the submission of the proposal, is the publication of a computational anatomical atlas of the left ventricle in Circulation, the leading journal in Cardiology, where I have clearly characterised the shape of the LV.
Once the anatomy is captured, a methodology is used to automatically uncouple the active relaxation and passive inflation of the ventricle, the two interrelated mechanisms that rule diastolic filling.
DIAGRAM: an overview of the model personalization technique is represented in this diagram.
MOVIE: clinically available data of deformation, captured through dynamic MRI, and pressure, measured with catheters, is assimilated into the model: the fundamental physical parameters of the model are optimised by minimising the differences between the predicted deformation by the model, and the observed deformation in images. The outcome of this process is the myocardial stiffness and relaxation profile that best explain the data. I have already provided the proof of concept of this methodology with the comparison of two diseased and one healthy subject, as published in the leading journal in the field of Medical Image Analysis.
2. On the other hand, the characterization of the atria-ventricular conduit will be tackled using one of my recent contributions, a method for the computation of blood pressure differences from velocity data captured by PC-MRI.
MOVIE 1: this imaging modality enables us to capture the velocity at each instant and voxel of the sequence. In this example we can see the streamlines of velocity colour coded by the magnitude of velocity.
MOVIE 2: solving the fundamental physical Navier-Stokes equations, the pressure that explains the acceleration and viscous friction of that velocity is computed. Now we have the same streamlines colour coded by pressure.
The idea is to characterise the presence of any impaired conduit function of the mitral valve through the existence of pressure drops.
I’d like to finish this presentation with the envisioned solution combining these two technologies, where the pressure that is required to estimate myocardial parameteres, currently only available through invasive catheterised procedures, is estimated with the non-invasive methods that have been described before. This will then lead to a simple clinical workflow for patients, in which they will only require an MRI acquisition for about 20 minutes, and where a computational modelling post-processing step unveils these novel biomarkers. The majority of patients could then benefit from this diagnostic tool minimising any associated risks, and therefore maximising the impact of this envisioned solution
Thank you for the introduction, and for the opportunity to present my ideas and goals.
The problem that I will address is the management of HF, a major health issue in the UK, which brings annual costs of £0.75 billion to our health system. The lifetime risk of developing HF is one in five, it is expected that 3 of us will develop some form of this disease.
HF is the clinical condition in which the heart is not able to pump enough blood to meet the body demands. Two actions govern the mechanical pump function of the heart, ejection and filling. The scope of my work focuses on the second, which relates to the condition of HFNEF, that affects half of the population with HF (the other half have systolic HF). HFNEF patients have abnormal filling, caused by a stiffer myocardium, a delayed relaxation, or an impaired atrio-ventricular counduit function.
Current diagnostic clinical guidelines for HFNEG use these surrogates to characterise an impaired filling of the ventricles: 1,2,3
The problem is that the characterization and stratification of patients is an on-going challenge. One of the fundamental reasons for it is that current diagnostic metrics are only surrogates of the mechanisms that impair ventricular filling. In my work, I plan to bridges this gap, estimating the fundamental mechanical properties that govern diastolic filling.