Automated segmentation and reconstruction of patient-specific cardiac anatomy and pathology fromin vivoMRI*

Abstract

This paper presents an automated method to segment left ventricle (LV) tissues from functional and delayed-enhancement (DE) cardiac magnetic resonance imaging (MRI) scans using a sequential multi-step approach. First, a region of interest (ROI) is computed to create a subvolume around the LV using morphological operations and image arithmetic. From the subvolume, the myocardial contours are automatically delineated using difference of Gaussians (DoG) filters and GSV snakes. These contours are used as a mask to identify pathological tissues, such as fibrosis or scar, within the DE-MRI. The presented automated technique is able to accurately delineate the myocardium and identify the pathological tissue in patient sets. The results were validated by two expert cardiologists, and in one set the automated results are quantitatively and qualitatively compared with expert manual delineation. Furthermore, the method is patient-specific, performed on an entire patient MRI series. Thus, in addition to providing a quick analysis of individual MRI scans, the fully automated segmentation method is used for effectively tagging regions in order to reconstruct computerized patient-specific 3D cardiac models. These models can then be used in electrophysiological studies and surgical strategy planning.

Effects of Fibrosis Morphology on Reentrant Ventricular Tachycardia Inducibility and Simulation Fidelity in Patient-Derived Models

“…Segmentation of cardiac images is a very challenging task because of the complex geometry and topology of the myocardium, movement artifacts from the acquisition process, blurred object boundaries in the images, and inhomogeneous image intensities between different voxel groups. 2 , 18 , 19 While expert interpretation is considered the gold standard, manual delineation of image boundaries within MRI scans is a time-consuming and tedious process, especially since each MRI dataset can contain between 10 to a few hundred images depending on acquisition protocol. It is also fallible, and subject to inter- and intra-observer variation.…”

“…To accurately automate the process, our group has engineered two automated algorithms to segment the LV 18 and the right ventricle (RV) 19 from cardiac MRI. These algorithms were utilized here to segment the ventricular myocardium at end-diastolic (ED) phase from the functional cine images for each patient.…”

Factors Promoting Conduction Slowing as Substrates for Block and Reentry in Infarcted Hearts

“…Here, we did not include prolonged APD along with reduced INa and other ionic changes (changes in L-Type calcium, for example) as has been implemented in more complete representations of the remodeled BZ in other studies (12,13,(15)(16)(17)(18)(19)(20)(21)(22) as our goal was to independently dissect the effects of individual remodeling processes. In such other work, more sustained episodes of induced arrhythmias were indeed witnessed.…”

“…Alternatively, it could also suggest that the balance between the specific functional wavelength relative to the anatomical (scar) re-entrant pathlength is more delicate in some species than others (i.e., once induced, particular species are more robust to slight differences in APD around the re-entrant circuit). This may also suggest why previous simulation work in human (18,21,22) and rabbit (13,15,17) chronic infarct models that have imposed similar functional remodeling (including prolonged APD) have not reported issues in simulating sustained re-entries. However, it is also interesting to note that a very recent work examining the re-entrant dynamics in similar ex vivo MRI-derived porcine models enforced a prolonged APD (and reduced CV) within BZ tissue only for arrhythmia induction (following the S1 beat); once induced, this was removed to facilitate arrhythmia sustenance, presumably due to issues in sustaining arrhythmias with such prolonged APDs (59).…”

“…APD prolongation can lead to the dispersion of repolarization between the BZ and the neighboring healthy myocardium, providing a known potent substrate for unidirectional conduction block (11). Indeed, computational studies from our group (12)(13)(14) and others (15)(16)(17)(18)(19)(20)(21)(22) have implemented repolarizing potassium channel remodeling to enforce prolonged APD in the BZ as a reliable means of facilitating unidirectional block around the scar. Such an in silico approach has been particularly valuable in helping to understand cardiac arrhythmic mechanisms and, more recently, move toward patient-specific simulation and therapy planning (23).…”