Abstract Details

Presented By:Haslam, Thom
Affiliated with:Comprehensive Arrhythmia and Research MAnagement (CARMA) Center, Cardiology, University Health Sciences
Authors:Thomas S. Haslam1, Joshua J.E. Blauer BS1,2, Nathan S. Burgon BS1, Yaw A. Adjei-Poku MD1, David Barlow1, Swati N. Rao1, Nassir F. Marrouche MD1, and Rob S. MacLeod PhD1,2
From:1 Comprehensive Arrhythmia and Research MAnagement (CARMA) Center, University of Utah School of Medicine., 2 Scientific Computing and Imaging (SCI) Institute, University of Utah.
Title
A Novel Image Processing Method to Calculate and Quantify Structural Remodeling of the Left Atrium
Abstract

Background. Atrial Fibrillation (AF) has been studied using delayed enhancement MRI (DE-MRI) to detect structural remodeling, which naturally occurs in the left atrium (LA) of patients with the disease. Custom MATLAB software has been used as the method for segmenting and quantifying structural remodeling. In this study, we present the use of Seg3D as a novel and efficient segmentation method.

Methods. For twenty-three patients who received a DE-MRI, the resulting scans underwent 2 segmentation-processing methods: first, using custom software written in MATLAB for manual boundary selection and then again, using the dedicated segmentation program Seg3D. Both processes supported contouring the epi- and endo-cardial borders of the LA and the resulting label map applied to the original image using custom MATLAB functions to calculate the amount of myocardial enhancement. A histogram of LA wall pixel intensities including the mean and standard deviation (SD) within this distribution are measured and intensities more than 3 SD above the mean were considered enhanced.

Results. The amount of myocardial enhancement was quantified using both segmentation methods and the result compared by correlation. The mean total LA enhancement using MATLAB was 16.4% ± 17.5% compared to the mean total enhancement using Seg3D, which was 11.3% ± 16.6%. A point-wise comparison of results showed a strong correlation with an R2 value of .754 (p<0.001). The statistical measure of inter-rater agreement (Kappa) for the study was .597. Seg3D was considerably faster and easier to use than the MATLAB software and supported a wider variety of image processing capabilities.

Conclusions. Seg3D provides an effective and reliable method to analyze tissue change in the LA and determine the amount of remodeling caused by AF. Knowing the extent of diseased myocardial tissue in a patient with AF will prove useful in determining treatment options and outcomes.