Segmentation of cardiac magnetic resonance cine images of single ventricle: including or excluding the accessorial ventricle?

Abstract

Our aim was to compare two different approaches for segmentation of single ventricle (SV) on cardiac magnetic resonance (CMR) cine images. We retrospectively studied 30 consecutive patients (23 males; aged 27 ± 10 years) with a treated SV who underwent 1.5-T CMR using ECG-triggered axial true-FISP, HASTE and cine true-FISP sequences. We classified patients for visceroatrial situs, cardiac axis orientation, ventricular loop, morphology of SV and position of great arteries. One experienced reader segmented cine images twice, firstly including only the systemic ventricle, secondly including both systemic and accessorial ventricles. Ejection fraction (EF), indexed end-diastolic volume (EDVI), end-systolic volume (ESVI), and stroke volume (SVI) were calculated. Data were presented as medians and interquartile intervals. Four patients presented dextrocardia and one patient mesocardia. Two had situs ambiguus with asplenia and one situs ambiguus with polisplenia. Four patients showed right morphology of the SV and three levo-ventricle loop. We found 14 levo-trasposition of great arteries (TGA), 4 levo-malposition of great arteries (MGA), four dextro-MGA, two dextro-TGA, and one inverted vessel position. When segmenting only the systemic ventricle, EDVI (mL/m²) was 65 (50–91), when segmenting both ventricles 76 (58–110) (P < 0.001); ESVI (mL/m²) was 32 (24–45) and 45 (33–60), respectively (P < 0.001); EF (%) was 49 (43–57) and 33 (24–47), respectively (P = 0.003); SVI (mL/m²) was 34 (17–48) and 33 (24–47) (P = 0.070). The inclusion of the accessorial ventricle in the segmentation of SV produce a biased lower EF showing a very low contribution to the pump function.