improving the stratification power of cardiac ventricular shape
TRANSCRIPT
Improving The Stratification Power
Of Cardiac Ventricular Shape
Gonzalez1, Nolte1, Lewandowski2, Leeson2, Smith3, Lamata1
1 Dept. Biomedical Engineering, King’s College Of London2 Dept. Cardiovascular Medicine, University Of Oxford3 Faculty Of Engineering, University Of Auckland
SUMMARY
Computational anatomy to improve shape stratification
INTRODUCTION &
OBJECTIVE
Motivation: measure cardiac shape remodelling
- Much more detail available in images
Hypothesis: Computational anatomy
METHODS
METHODOLOGY
Capture anatomy in a consistent manner
- Mapping, correspondence…
Reduce dimensionality (statistics)
…
IN MORE DETAIL…
1. Mesh personalization [1,2]
2. Atlas construction: mean + anatomical modes (Mi)
[1] Lamata et al. “An automatic service for the personalization of ventricular cardiac meshes.” J R
Soc Interface. 2014
[2] Lamata et al. “An accurate, fast and robust method to generate patient-specific cubic Hermite
meshes.” Med Image Anal. 2011
-2std +2std
ANATOMICAL MODE
SHAPE COEFFICIENTS
The directions of shape change
- Mathematically perfect, capturing biggest variance or differences
- Clinically difficult to interpret
How much of change in each direction
(each anatomical mode)
Shape = mean + Sum (Ci * Mi)
Coefficient
Anatomical Mode
3 CASE
STUDIES
CASE 1: PREDICT
GESTATIONAL AGE (I)
Study of effect of premature birth
- Adults (20s to 30s)
- Subgroups: pre-term (30±2.5 weeks), term birth (40±1 weeks)
Circulation. 2013 Jan 15;127(2):197-206.
CASE 1: PREDICT
GESTATIONAL AGE (II)
Circulation. 2013 Jan 15;127(2):197-206.
CASE 1: PREDICT
GESTATIONAL AGE (III)
5 clinical metrics:
- Length
- Epicardium diameter
- Endocardium diameter
- Cavity volume
- Mass
Computational mesh Modes of variation
Classification
Task
Conventional metrics
Input images
CASE 1: PREDICT
GESTATIONAL AGE (IV)
CASE 2: REVEAL HLHS
REMODELLING (I)
Hypoplastic Left Heart Syndrome (HLHS)
Reveal impact of shunt choice
MBT: Modified Blalock-Taussig
RVPA: Right Ventricle to Pulmonary Artery
CASE 2: REVEAL HLHS
REMODELLING (II)
Ventricle grow differently depending on surgical choice in
HLHS [M12].
[M12] Wong et al. “Using Cardiac Magnetic Resonance and Computational Modelling to Assess
the Systemic Right Ventricle Following Different Norwood Procedures: A Dual Centre Study”
CASE 3: PREDICT AF
RECURRENCE (I)
Problem: atrial fibrillation recurrence after ablation
Shape of the left atrial blood pool to predict recurrence
Antero-Posterior direction
S
I
LR
Average recurrent
Average non-recurrent
CASE 3: PREDICT AF
RECURRENCE (II)
Second mode: better predictive power than previous metrics
(work in progress)
Generate virtual extreme geometries within the range of
physiological variation
Antero-Posterior direction
S
I
LR
Extreme recurrent
Extreme non-recurrent
CONCLUSIONS
CONCLUSIONS
Shape is much more than length or volume
Computational Anatomy tools mature and available
http://amdb.isd.kcl.ac.uk/
Disclaimer: research prototype, easily adaptable to needs,
but be patient if not 100% reliable!