Project Description
Most cardiovascular diseases impair the normal motion of the left ventricle (LV). Recent improvements in tomographic imaging devices such as MRI or CT, have led to the creation of 3D computer models that show the dynamics of the LV. Since the mechanical properties of any muscle are determined by the arrangement of their fibres in space, a model that aims to realistically simulate the behavior of this tissue must take into account the spatial distribution of the fibres. In the case of the myocardium it plays a crucial role for the understading of the electromechanical activation sequence in normal and pathologic conditions and the mechanisms of the ventricular resynchronization therapy (pacemaker implantation).
The MIOCARDIA project is a multidisciplinary framework in collaboration with the Hospital de la Santa Creu i Sant Pau, Clínica La creu blanca and the Barcelona Supercomputing Center. Its purpose is to provide an integrative computational model of the function and anatomy of the heart.
PARTNERS
Hospital de Sant Pau, Universitat de Lleida, Barcelona Supercomputing Centre
FUNDING
TIN2009-13618
Innovative aspect
Current commercial software offer fewer advantages than HPF, as they don’t correctly model movement at late stages of systole. HPF is a simple algorithm (it can be computed by solving a linear system) which incorporates regularization of the vector field only at image areas of poor quality. This results in motion estimation vector fields properly modelling tissue motion provided that the image quality allows its visual assessment. From the clinical point of view this implies that motion is not overestimated at injured (pathological) areas. From the image processing point of view, the algorithm can be easily adapted to other image sequences to extract more reliable motion fields.
NPD bases on parametrizing the LV domain for (implicitly) registering images acquired for different patients and devices, which endows NPD with several advantages over existing techniques. The unit cube defining NPD is a novel framework for easily moving on the LV in (natural) angular and radial coordinates and allows straightforward comparison and fusion of LV scores obtained from different image modalities.
Normality models for (complementary) LV integrity scores (including stains and 2D motion) have been computed. We have explored the performance of several sets of functional descriptors and provided with the necessary (statistical) tools for deciding which configuration is the best suited for pathology discrimination.
Main advantages
By working in frequency (Gabor) domain HPF formulation is given by simple equations (opposite to registration algorithms working in intensity domain). The use of Gabor filters makes HPF able to properly model local tissue deformation at advance stages of the systolic cycle (in contrast to current techniques relying on Fourier). By applying the regularization term only at poorly tagged areas we do not overestimate motion at injured hypo kinetic territories.
NPD provides a generic normalized domain (a unit cube) for comparison and merging of functional scores extracted from different image modalities and, thus, it is well suited for computation of multi-parametric normality ranges. Definition of intuitive (for physicians) coordinates (radial and angular) becomes straightforward, as they correspond to cartesian coordinates (matrix rows and columns) in the unitary cube.
We have defined a statistical strategy which allows designing the set of clinical scores that best characterize and discriminate a specific pathological group.
Advantatges
Help in diagnosis and planning of cardiovascular diseases treatments, especially in cardiac resynchronization on (pacemaker) therapy , where a proper recovery of a normal LV function is not achieved in many occasions.
The methods developed might constitute a valuable tool for checking LV anatomic and functional theories and models, since they provide the experimental ground truth data for tagged magnetic resonance
A tangible result is our plugin for the Osirix Platform, capable of extracting clinical scores from Tagged Magnetic resonance DICOM series.