Abstract 9089: Impact of Semiautomated Border Tracking on the Analysis Time and the Accuracy of Right Ventricular Volume Measurements Using DICOM Gradient Cine MR Datasets
Background: Right ventricular (RV) volume measurements on MR images require significant time for tracing multiple short axis planes and is subject to errors due to through plane motion of tricuspid plane, which limits routine clinical application. We tested a new 3D method of RV volume estimation using semiautomated border tracking to reduce both errors and analysis time.
Methods: We studied 20 latex RV shaped models of different sizes (25-225ml). Each RV model was driven by a calibrated pulsatile pump at 3-5 different stroke volumes (SV, 20-80ml). Motion of the RV model was scanned for gradient cine stacks in multiple short axis, sagittal and coronal planes with a 3T Siemens MR Scanner. Image data was analyzed on an offline RV analysis program embedded in Research Arena VA® (TomTec). Each cine loop was analyzed with manual tracing employing automated border tracking. Datasets were then analyzed for RV volume after manual editing of borders at each short axis level.
Results: Semiautomated tracking used 3-5 minutes rather than the manual 30-45 minutes. Both methods overestimated ESV and underestimated SV but showed a positive linear correlation (r = 0.80 & 0.87 for ESV and 0.85 & 0.89 for SV respectively) compared to reference values, and good interobserver variability (r = 0.92 & 0.77 respectively). Bland-Altman analysis showed slightly lower bias for manual editing (3.45±0.80) compared to automated tracking (5.25±1.45) but all data points were within 95% confidence intervals.
Conclusions: The new RVMR program is a user friendly semiautomated program that may be used for clinical assessment of RV volumes.
- © 2012 by American Heart Association, Inc.