The aim of the following experiment is to estimate fiber orientations of a patient heart for personalized simulations of cardiac electrophysiology because it is not currently feasible to clinically acquire cardiac fiber orientations. This is achieved by deforming an atlas heart geometry to match the geometry of the patient heart available from a clinical CT or MRI image as a second step. The geometry deformation field is used to morph the atlas fiber orientations, which then generates an estimate of the patient heart fiber orientations.
Next, using canine hearts whose fiber orientations are measured non clinically using diffusion tensor, MRI, the state-of-the-art technique, the estimated fiber orientations are compared with acquired ones in order to validate the proposed methodology. Results obtained show that the estimated fiber orientations match closely with acquired ones and there are no significant differences between outcomes of electrophysiological simulations with estimated and acquired fiber orientations at a clinically observable level. This research demonstrates quantitatively that in the absence of diffusion tensor, MRI myocardial fiber orientations of normal and failing ventricles can be estimated from in vivo images of their geometries for use in simulations of cardiac electrophysiology.
Estimation of the patient heart fiber orientations is essential to our ability to construct patient specific models of heart function. These patient specific heart models can be used to predict the risk of arrhythmia in patients with heart disease or to guide interventions that can be used to treat disorders in heart rhythm. Although we demonstrate that this can be used for patient specific electrophysiological modeling, it can also be applied for personalized electromechanical modeling.
Begin to estimate the fiber orientation of a subject's heart by first acquiring structural MRI and diffusion tensor MRI or D-T-M-R-I. Images of a normal adult human heart in diastole. Acquire the images at a resolution of one millimeter cubed.
Then use image J to extract the ventricular myocardium from the atlas structural image. Do this by fitting close blinds through a set of landmark points along the epicardial and endocardial boundaries. For each short axis slice, perform the placement of landmark points manually.
For every 10th slice in the image, obtain the landmark points for the remaining slices by linearly interpolating the manually identified points using matlab. Next, reconstruct the fiber orientations of the atlas heart by computing the primary iGen vectors of the diffusion tensors in the DT MRI image. Once the geometry of the atlas heart has been defined, acquire an image of the geometry of the patient heart in diastole using in vivo cardiac CT or MRI reconstruct the patient heart geometry from the image similarly to the way the atlas was built.
Ensure that the resolution of the reconstructed geometry is one millimeter cubed by adjusting the number of slices for which the landmarks are manually picked and the interval of out of plane interpolation. Next, deform the atlas ventricular image shown in magenta to match the patient geometry image shown in red in two steps. In the first step, perform an aine transformation based on a set of 13 landmark points described in the protocol.
In the second step, further deform the INE transformed atlas ventricles to match the patient geometry using large deformation dimorphic metric mapping or L-D-D-M-M. Next morph the DT MRI image of the atlas. To accomplish this, reposition the image foxholes and reorient the diffusion tensors.
According to the transformation matrix acquired from the AINE matching and the deformation field during the L-D-D-M-M transformation, the reorientation of the diffusion tensors should be performed using the preservation of principle directions or the PPD method. Finally, obtain an estimate of the patient fiber orientations from the morphed atlas GT MRI image by computing the primary iGen vector of the diffusion tenses. In order to measure the estimation error, first acquire ex vivo structural MRI and DT MRIs of six normal and three, failing canine hearts at high resolution.
Segment the ventricles from the canine hearts similarly to those of the human atlas heart, as described in the previous section, denote ventricles segmented from normal canine hearts as hearts one through six, and those from failing canine hearts as heart seven through nine. Then use each of hearts two to six as an atlas to obtain five different estimates of ventricular fiber orientations of heart. One next estimate fiber orientations for each of the failing ventricles from heart seven, eight, and nine, using heart one as the atlas for all of the data points in each set of estimated fiber, orientations compute the estimation error where theta E and theta A are the inclination angles of estimated and acquired fiber orientations at that point respectively from heart.
One construct six models construct the first model with the DT MRI acquired fiber orientations of heart one and models two through six with the estimated fiber orientations, datasets of heart, one for each of the three. Failing heart geometries construct two ventricular models, one with the DT MRA acquired fiber orientations and the other with the estimated fiber orientations. Here, the spatial resolution of the models should be 600 microns.
Next signify the heart failure models. With DT MRI acquired fibers as models seven to nine and those with estimated fibers as models 10 to 12 in the models. Use mono domain representation to describe the cardiac tissue by utilizing the governing equation shown here and described in detail in the text protocol using the software package carp by cardio solve, simulate sinus rhythm with all models using PACE insights as shown here, overlaid on heart seven.
Next, induce reentrant ventricular tachycardia in the six failing models using an S one S two pacing protocol with electrode locations as illustrated here to accomplish this, choose the timing between S one and S two to obtain sustained ventricular tachycardia activity for two seconds after S two delivery. If ventricular tachycardia is not induced for any S one S two timing, decrease the conductivities by up to 70%until tachycardia is induced for each simulation. Calculate pseudo ECGs by taking the difference of extracellular potentials between two points near the base of the heart.
The points are separated by 18 centimeters such that the line connecting them is perpendicular to the base apex plane of the septum as illustrated by E one and E two for each simulation with the estimated fiber orientations compute the mad metric using the formula shown here where X is the ECG waveform of estimated fiber orientations. Y is the ECG waveform of the acquired fiber orientations. X minus is the mean value of x.
Y minus is the mean value of Y, and then is the length of X and Y Shown here are streamlined visualizations of a healthy and diseased heart shown with estimated fiber orientation shown as cyan fibers and DT MRI acquired fiber orientations shown in yellow. Looking more closely, you can see the similar alignment between the acquired and estimated fiber orientations. The difference in inclination angles between the acquired and estimated fibers is more easily seen in a 3D distribution model that uses color to represent the mean estimation error angle.
The color bar ranges from zero degrees to 175 degrees from this data. A histogram showing the number of voxels at each degree of error is plotted for both the normal and failing heart. These results show that the inclination angles of predicted fiber, orientations of normal and failing hearts are comparable to those acquired by ex vivo DT MRI.
The simulated activation maps are then used to compare the acquired and estimated fiber orientations with respect to the electrical signal propagation throughout the heart. The color represents activation time as described by the color legend. The activation map on the left is for simulation of sinus rhythm with model one representing the acquired fiber orientations, and those on the right are for simulations of sinus rhythm.
With models two through six representing estimated fiber orientations. When the activation maps corresponding to the estimated fiber orientations are averaged the mean overall difference between them and model one is minimal. The pseudo ECGs of sinus rhythm simulations with model one and model three show identical morphologies and result in a low MAD score of just 4.14%Similarly, when comparing activation maps of sinus rhythm simulations with acquired and estimated fiber orientations of the diseased hearts, low MAD scores were seen.
The MAD score ranged from 3.8 in heart, seven to 6.1 in heart nine. This represents a high correlation between the acquired and estimated fiber orientations Once mastered, this technique can be executed in a few hours if performed properly Following this procedure. Estimation of fiber orientations can be performed in hearts with ischemic cardiomyopathy in order to test the efficacy of the method in the presence of myocardial infarction.
The development of this technique paves the way for translation of computer simulations of heart function from basic science into the clinic where they can be used at patient bedside to guide interventions for treating disorders in heart rhythm.
