2012 Progress Report DBP Atrial Fibrillation

Jump to: navigation, search
Home < 2012 Progress Report DBP Atrial Fibrillation

April 18, 2012 CARMA DBP NAMIC Progress July 2011 – June 2012

Section A: Introduction

Tissue remodeling of the atrial wall is the hallmark of Atrial Fibrillation (AF), a progressive cardiac disease that develops over time (months to years). The mechanisms of this transformation are only partially understood, but the current scientific focus on tissue remodeling and its putative role in AF suggests that novel forms of MRI can be used to evaluate new patients, predict success before ablation, analyze outcomes post-ablation, and guide repeat ablations. Such MRI-based therapies, however, urgently require advanced tools and software to support efficient workflows and accelerate the quantification and analysis of images. The CARMA-NAMIC DBP project is addressing these needs through the development of algorithms and tools for the automated segmentation of heart structures and the MRI-based evaluation of AF progression and recovery.

The CARMA-NAMIC DBP research and development activities are organized under three Specific Aims as follows:

1. Develop and validate image-based longitudinal diagnostic indices for AF 2. Develop automatic segmentation methods for the left atrium (LA) and adjacent structures 3. Develop an AF scoring scheme to evaluate disease progression and recovery from therapy

Section B: Research and Progress Report

CARMA-NAMIC DBP activities over the past year include tangible progress towards Specific Aim 2, as we continue to work closely with collaborators from the Algorithms core to develop strategies for automated segmentation of the left atrium from LGE-MRI. Yi Gao (BWH) has proposed a multi-atlas segmentation method for the LA endocardium that we have tested on our clinical data and are helping to refine to be more robust. We are also providing data and evaluating results for a project with Liangjia Zhu (Georgia Tech) to evaluate a novel approach to active contour segmentation that includes an automatic identification of seed regions within the LA. More recent is a very active collaboration with Ross Whitaker (Utah) to test and refine a novel, globally-optimal graph cut algorithm for LA wall segmentation.

In support of these, and other, collaborations, we released a public LGE-MRI database of 33 AF patient scans at pre- and post-ablation, along with matching MRA data and manual LA segmentations. We also conducted an internal study of the consistency and accuracy (compared to a probabilistic ground truth) of the manual LA segmentations produced by the technicians at CARMA, to serve as a basis for comparison with automated approaches.

Another focus of DBP activities this year was the development of robust and practical image registration algorithms, which are important for the accomplishment of all DBP Specific Aims. One important research area is the longitudinal analysis of tissue changes in the LA wall and the pulmonary veins, for which we have identified several common clinical use-cases and, in collaboration with Yi Gao (BWH), have worked out practical pipelines and parameter settings that we are currently implementing in Slicer 4. We also have preliminary results for the registration of previously acquired MRI images and scar maps into the imaging space of fluoroscopy-guided ablation procedures. Merging of MRI based, patient specific information into the ablation will provide clinicians with a novel means of integrating information with spatial fidelity that they can now only combine in a very qualitative manner.

A major highlight of our work this year on diagnostics and clinical evaluation (Specific Aims 1 and 3) was our development of a more robust automatic approach to post-ablation LA scar segmentation. We presented this new method at SPIE Medical Imaging in February 2012 and are currently implementing it as a Slicer 4 module. We are now applying similar approaches to create a robust algorithm for fibrosis segmentation, a much more challenging problem because of the subtle variations in image intensity associated with fibrosis. Initial results have proven accurate when compared to manual ground-truth segmentations. In May 2012, we will participate in a scar and fibrosis segmentation challenge at the IEEE Symposium on Biomedical Imaging, which we co-organized with Kawal Rhode at Kings College, London, and Dana Peters at Yale. The goal of the challenge is to highlight this particularly challenging segmentation problem and support quantitative evaluation of algorithms and then discussion of their relative performance.

Section C: Plans for the Coming Year

Research activities in the next year include further development of automatic segmentation of the LA, through collaboration with Ross Whitaker and the GA Tech group, and refinement and testing of our registration techniques for both diagnostic (LGE-MRI to LGE-MRI) and therapeutic (LGE-MRI to fluoroscopy) purposes. We will also continue to focus heavily on developing a more automated approach to segmentation of fibrosis in preablation images. While our current approach compares well with manual ground truth, it still requires improvements in sensitivity to achieve the robustness of manual methods, especially in predicting successful outcome of RF ablation (the Utah AF scoring scheme). We also plan to release additional patient data to supplement the existing public database. Specifically, we will include not only more subjects but also additional time points (6-month and 1-year follow-up scans) and anatomical landmarks (pulmonary vein, mitral valve, and appendage attachment locations).

Over the next year we will increase our engineering activities to disseminate algorithms from the CARMA-NAMIC DBP through Slicer 4. Specifically, we will release 1) a registration module for cardiac LGE-MRI that includes preset parameters for specific use-cases 2) an intensity inhomogeneity correction module 3) a parameter-free automatic scar segmentation module and 4) a landmark-based registration module for pre/post pulmonary vein antrum evaluation. We will also work closely with our algorithms partners to test and develop their Slicer 4 automatic segmentation modules.

Section D: Papers that Acknowledge NAMIC

Automatic Segmentation of the Left Atrium from MRI Images using Salient Feature and Contour Evolution Liangjia Zhu, Yi Gao, Anthony Yezzi, Rob MacLeod, Joshua Cates and Allen Tannenbaum EMBS 2012, submitted

Automatic classification of scar tissue in late gadolinium enhancement cardiac MRI for the assessment of left-atrial wall injury after radiofrequency ablation Daniel Perry, Alan Morris, Nathan Burgon, Christopher McGann, Robert MacLeod, Joshua Cates SPIE Medical Imaging 2012