Abstract 729: A More Accurate Method for Automatic Segmentation of Left Ventricular Mass by SPECT

Abstract

Background: Accurate quantification of left ventricular mass (LVM) by single photon emission computed tomography (SPECT) is important when analysing LVM as such, but also when estimating the relative size of a perfusion defect. Existing algorithms for segmentation of LVM by SPECT have undergone limited validation. The aim of this study was to develop and validate a new segmentation algorithm for LVM by SPECT, and compare the new method to the commercially available software Quantitative Gated SPECT (QGS). Methods: 98 patients with known or suspected coronary artery disease underwent both rest and stress SPECT and MR. The new method was trained in 20 patients (40 studies) and tested in 78 patients (156 studies) by comparing LVM in both rest and stress SPECT with manually segmented LVM by magnetic resonance imaging (MRI) as the reference standard. The new method was compared to results from QGS. Results: The mean difference ±SD in LVM between SPECT and MRI was lower for the new method (6±15 %LVM) compared to QGS (18±19 %LVM) for both mean difference (p<0.001, t test) and SD (p=0.01, F test), see Figure . Conclusion: The new method quantifies LVM with a significantly lower bias and variability compared to the commercially available QGS software, when compared to manually segmented LVM by MRI.