Abstract 10381: Improved Left Ventricular Mass Quantification with Partial Voxel Interpolation - in vivo and Necropsy Validation of a Novel Algorithm for Automated Analysis of Cardiac Magnetic Resonance (CMR)
Background Partial voxel analysis, a new segmentation method for LV mass (LVM), accounts for myocardium admixed with blood in a single voxel. This differs from conventional full voxel analysis, which partitions regions in a binary manner based on manual or automated contours and may assign voxels containing blood and myocardium to LV cavity. This is the first study to test partial voxel segmentation for LVM.
Methods CMR was done in patients and lab animals: LVM was measured by 3 methods - manual planimety (MP), automated partial voxel (ASPV), and full voxel (ASFV) segmentation. Patient results were compared to LVM by a necropsy validated echocardiography method (echo). Lab animal results were compared to LV weight at necropsy
Results 126 patients with CAD and 10 lab animals (8 dogs, 2 pigs) were studied.AS quantified LVM in all patients, yielding a 12 fold shorter processing time than MP (0:21±0:04 vs 4:18±1:02 min, p<0.001). ASFV (136±35gm) was slightly lower than MP (139±35; Δ=3±9gm, p<0.001) though both yielded similar proportions of patients with LV remodeling (56% vs 57% p=.7) and hypertrophy (4% vs 3% p=1.0). Regarding partial voxel segmentation, ASPV yielded higher LVM (159±38gm) than MP (Δ=20±10gm) and ASFV (Δ=23±6gm, both p<0.001), reflecting 14% and 17% increases. Differences between ASPV and ASFV independently correlated with larger voxel size (r=.37, p<0.001), LV volume (r=.28, p=0.002), and LVM (r=.19, p=0.03). In patients (Figure A), ASPV yielded better agreement with echo (Δ=20±25gm) than did ASFV (Δ=43±24gm) or MP (Δ=40±22gm, both p<0.001). In lab animals (B) ASPV and ex vivo results were similar (Δ=1±3gm, p=0.3), whereas ASFV (6±3gm, p<0.001) and MP (4±5gm, p=0.02) yielded small but significant underestimate of LVM at necropsy
Conclusions Compared to conventional methods of MP and ASFV, partial voxel analysis yields improved CMR agreement with independent references of LVM measured at necropsy in lab animals, and a necropsy validated echo method in humans.
- © 2011 by American Heart Association, Inc.