Abstract 13889: Fully-automatic Segmentation Algorithm Monitors Left Ventricle Myocardial Remodeling Accurately in Patients With Heart Failure and Resynchronization Therapy With CT Angiography
Introduction: The Left ventricular myocardium (LVM) can be assessed using a model-based automatic algorithm with CT Angiography. We aimed to evaluate the accuracy of this segmentation algorithm in monitoring the myocardial remodeling in heart failure patients with CT angiography (CTA).
Hypothesis: We assessed the hypothesis that the model-based automatic algorithm can accurately assess the cardiac volume in the heart with the various geographic shape and phases of RR interval.
Methods: This study consists of 63 patients (35 female, mean 61yrs) with heart failure and cardiac resynchronization therapy, who underwent both baseline and follow up retrospective CTA in one-month interval. The LVM was measured using a manual method (GE ADW system, USA) for images at 35%, 75% and 95% of RR interval, as a reference, and fully automatic assessment (Philips EBW system, Netherlands) on all 20 phase images of RR. The coefficient variances between individual LVM (LVMi) of each phase and mean value (LVMm) of 20 phases were calculated (CV, in %). The LVM change rate [(a-b) /a х 100%] between baseline and follow-up of one month was compared among the manual and automatic assessment.
Results: Overall, in 63(100%) and 51 (81%) cases, the LVM was successfully measured by using the manual and automatic segmentations respectively. The expended time was 15 minutes for manual and 2 minutes for automated. A considerable linear correlation between both segmentation methods (r= 0.89, p<0.001) was found. There was significant constant within the LVMi and LVMm in 3.0 ±1. 5% on average of 20 phases assessments (r=0. 98-0.99). Using 6% (Mean +2Sd) as the cutoff point for the marked LVM change, the LVM was increased in nine cases (6.1- 22%), no change in 27 cases (-6 to 6%) and a decrease in 15 cases (-6.1 to -24.3%) with manual measurement. Comparing to the manual method in LVM change rate, agreement of automatic segmentation was in 38 of 51 (74.5%) cases with a significant association between both methods (r=0. 70, P<0.001).
Conclusions: In conclusion, myocardial remodeling in patients with heart failure can be monitored precisely and accurately using a model-based automatic segmentation with EBW system and CTA.
- © 2013 by American Heart Association, Inc.