Magnetic resonance imaging--cardiac ejection fraction measurements. Phantom study comparing four different methods.

Journal Article (Journal Article)

The accuracy of cardiac ejection fraction (EF) measurements with thin, contiguous cine-magnetic resonance imaging (MR) sections is well established. Still, faster imaging and measurement techniques would be desirable. The authors evaluated the accuracy of four different MR EF measurements methods in a biventricular, anthropomorphic, foam-latex rubber phantom which was connected via noncompliant fluid-filled tubing to a pulsatile flow pump. Nine contiguous 10 mm cine-MR sections (TR/TE, 25/13; flip angle, 45 degrees) were obtained through the heart in long and short cardiac axes at 16 frames per cardiac cycle at a pump rate of 60 beats/minute. EF measurements were based on either the multi-slice summation technique (nine contiguous 10-mm sections versus four 10-mm sections spaced 10 mm apart) or the area-length method (single largest long section versus combination of largest long- and short-axis section). Three replications were performed for each of the tested EFs (40.8%, 29.4%, and 13.4%), which were compared with actual EFs. EF measurements based on contiguous 1-cm sections correlated best with the actual EFs. Average relative errors ranged from 3.2% to 6.0%. EF measurements based on every other section were less accurate; average relative errors were between 5.2% and 10.2%. Single and biplane area-length algorithm EF measurements were significantly less accurate; average relative errors were as high as 59%. EF measurements based on multi-slice summation are more accurate than those based on the area-length algorithm. Contiguous 1-cm section acquisitions are most accurate and most time consuming. With slight decrease of accuracy, acquisition and processing times can be halved by skipping every other slice.

Full Text

Duke Authors

Cited Authors

  • Debatin, JF; Nadel, SN; Sostman, HD; Spritzer, CE; Evans, AJ; Grist, TM

Published Date

  • March 1992

Published In

Volume / Issue

  • 27 / 3

Start / End Page

  • 198 - 204

PubMed ID

  • 1551769

International Standard Serial Number (ISSN)

  • 0020-9996


  • eng

Conference Location

  • United States