Automatic Computation of Ejection Fraction using Spatio-Temporal Information
Pednekar A, Flamm S, Cheong B, Muthupillai R
Philips Medical Systems, St.Lukes Episcopal Hospital and Texas Heart Institute
The current clinical practice of manually tracing the endocardium to compute the left ventricular (LV) ejection fraction (EF) from cine-magnetic resonance images (cine-MRI) is labor intensive, time consuming, and operator dependent. We used spatio-temporal information in the cine-MRI to automatically estimate the EF in 64 subjects (21 volunteers, and 43 patients). This entirely data-driven, multi-step approach incorporates apriori geometric information from the acquisition and spatial-temporal information intrinsic to cardiac cine-MRI. The results show that the mean bias values determined by Bland-Altman analysis for end-diastolic and end-systolic volumes and hence for EF are in close agreement with inter- and intra-observer variability.