733 Views
Oral Abstract Session
SCMR 22nd Annual Scientific Sessions
Bharath Ambale-Venkatesh, PhD
Instructor
The Johns Hopkins University
Mounes Aliyari ghasabeh, MD
post doctoral research fellow
John Hopkins Hospital
Anneline te Riele, MD, PhD
cardiologist
john hopkins hospital
Cynthia James, PhD
Assistant Professor
john hopkins hospital
Crystal Tichnell, PhD
Genetics Counselor
john hopkins hospital
Brittney Murray, MD, PhD
Genetics Counselor
john hopkins hospital
Ela Chamera, PhD
Research Sonographer
Johns hopkins
Joao Lima, MD
Professor of Medicine, Radiology and Epidemiology
Johns Hopkins University
Birgitta Velthuis, MD
Professor of cardiology
UMC Utrecht
Harikrishna Tandri, MD
Associate Professor
john hopkins hospital
Hugh Calkins, MD
PROFESSOR
Johns Hopkins Medicine
Ihab Kame, MD, PhD
PROFESSOR
john hopkins hospital
Stefan Zimmerman, MD
ASSOCIATE PROFESSOR
The Johns Hopkins University
Background:
Arrhythmogenic Right Ventricular Dysplasia/ Cardiomyopathy (ARVD/C) is characterized by fibrofatty replacement of cardiac muscle and lethal arrhythmias. Diagnosis is based on 2010 Task Force Criteria (TFC). CMR is included in the TFC and wall motion abnormalities are required to meet CMR criteria. However, identification of wall motion abnormalities in the right ventricle (RV) is highly subjective. Feature tracking CMR is used to quantify myocardial wall motion (strain) and may be a solution to overcome subjectivity in wall motion abnormality detection. One difficulty in the use of strain techniques is the large amount of regional and global strain estimates and the associated challenge in identifying which estimates are most relevant to diagnosis. Our purpose, therefore, was to investigate the value of CMR-based strain analysis to diagnose ARVD/C in a cohort of suspected ARVD/C patients.
Methods: This retrospective, multi-center, international study included 395 consecutive ARVD/C-suspected patients who had been referred to two tertiary centers in the United States and the Netherlands for ARVD/C evaluation. All patients had a CMR. After a complete TFC work-up, patients were stratified into 3 groups: 1: definite ARVD/C, 2: at-risk, 3: non-ARVD/C. For all patients, RV and left (LV) ventricular strain analysis was performed on cine SSFP CMR images using Myocardial Tissue Tracking software. We measured 50 variables including global and segmental RV/LV strains, volume, function, and demographic information. Random forest classifiers were used for identifying the ability of the variables to classify patients into ARVD/C and non-ARVD/C groups. Selection of the most effective classifier variables was performed using the Gini-index measure of variable importance. ROC analysis was performed to evaluate accuracy of the important variables to detect ARVD/C and at-risk from non-ARVD/C patients.
Results: The variable importance based on Random forest classifiers is shown in Figure 1. In both comparison of ARVD/C to non-ARVD/C, and at-risk to non-ARVD/C patients, basal and global RV longitudinal strain measures had the highest importance and were the most effective classifiers. ROC analysis showed global RV longitudinal strain alone has the sensitivity of 84%, specificity of 81% and accuracy of 87% for ARVD/C detection. For identification of at-risk patients, global RV longitudinal strain had a sensitivity of 70%, specificity of 70% and accuracy of 77%.
Conclusion: Among various myocardial strain parameters, global RV longitudinal strain was the most useful for diagnosing ARVD/C, and had greater importance for predicting the diagnosis as compared to traditionally used RV volumes and function from CMR. Quantitative myocardial strain analysis showed improved sensitivity, specificity and accuracy in detecting ARVD/C patients. Further investigation into whether this method could be a solution to overcome subjectivity in detecting wall motion abnormality in the current TFC is warranted.