Amol Pednekar1, Mercedes Pereyra2, Brenda Lambert3, Debra Dees2, Benjamin Cheong2,4, Raja Muthupillai, 2,5
1Philips Healthcare, Houston, TX, USA; 2Department of Radiology, St.Lukes Episcopal Hospital and Texas Heart Institute, Houston, TX, USA; 3Department of Radiology,, St.Lukes Episcopal Hospital and Texas Heart Institute, Houston, TX, USA; 4Department of Cardiology4, St.Lukes Episcopal Hospital and Texas Heart Institute, Houston, TX, USA; 5Department of Radiology, Baylor College of Medicine, Houston, TX, USA
Manually delineation of the endocardial contours to compute left ventricular (LV) volumes and ejection fraction (EF) from cine-magnetic resonance images is labor intensive, and operator dependent. We clinically validated an automatic algorithm that computes the LVEF using classification of partial volumed pixels based on temporal intensity information in conjunction with geometrically smooth parametric curve on 16 volunteers (LVEF mean 58%, range 49-70%) and 7 patients (LVEF mean 53%, range 40-65%). The mean bias values computed with Bland-Altman analysis (EF-2%, EDV-2ml, ESV-4ml) between automated contours and expert manual contours is comparable to the typical inter- and intra-observer variability of experienced clinicians.