103: Surface-Length Index (SLI): a novel index to predict right ventricle dysfunction by cardiac MRI

Abstract

Short axis CineMRI has become the gold standard to measure Right Ventricular Ejection Fraction (RVEF). However, it requires a time-consuming manual contouring of the endocardium. Therefore, many examinations do not include complete RV study. We hypothesized that a simple index could be used to detect patients with abnormal RVEF requiring a precise RV study. Two classical RV function indices were tested: 1/ RV fractional area change (FAC) measured in a mid-ventricular short-axis slice and 2/ RV shortening fraction (SF) corresponding to the longitudinal contraction measured in the horizontal long-axis view. They were then combined to obtain the Surface-Length Index: SLI=1- (1-FAC) (1-αSF). This formula was derived from the application of a basal-to-apex gradient in the RV longitudinal contraction into the crescentic shell model published in 1989 by Aebischer. 400 patients underwent a conventional cardiac MRI with a horizontal long-axis view and a stack of contiguous 8 mm short-axis slices at 1.5T with SSFP sequences and were divided in 2 groups: 60 patients retrospectively included to determine α by optimization of the area under ROC curves (Group A) and 340 patients prospectively included to test SLI, FAC and SF capacity to predict a RVEF alteration (<0.5) ( Group B). In group A, the optimal value for α was 1.3. In group B, SLI, FAC and SF area under the ROC curves were respectively 0.94, 0.87 and 0.81. SLI allowed a better detection of RV dysfunction (p<0.001). With a threshold of 0.50, SLI yielded sensitivity and specificity of 85 and 89%. With a threshold of 0.58, SLI yielded sensitivity of 100% and specificity of 51%. Using SLI with a threshold of 0.58 would have saved 38% of the operator's time without inducing any diagnostic error. SLI/EF correlation was good (r 2 =0.64) (cf figure). SLI combines two simple RV measures in end-diastole and end-systole (around 1 minute), and allows significant improvement in post-processing efficiency by pre-selecting RV requiring a complete study. The gain in the operator processing time is close to 1/3. Figure Abstract 103 – Correlation between SLI (vertical) and EF (horiz.)