Skip to main content
EJNMMI Research

Table 2 Washout rates determined from SPECT and planar images using CNN-based and standard methods, respectively

From: Convolutional neural network-based automatic heart segmentation and quantitation in 123I-metaiodobenzylguanidine SPECT imaging

  Planar WRBC Total
< 34% ≥ 34%
(A)    
SPECT WRCNN    
 ≤ 30% 23 (48.9%) 1 (2.1%) 24 (51.1%)
 > 30% 5 (10.6%) 18 (38.3%) 23 (48.9%)
Total 28 (59.6%) 19 (40.4%) 47
  Planar WRNC Total
< 30.1% ≥ 30.1%
(B)    
SPECT WRCNN    
 ≤ 30% 18 (38.3%) 1 (2.1%) 19 (40.4%)
 > 30% 9 (19.2%) 19 (40.4%) 28 (59.6%)
Total 27 (57.5%) 20 (42.5%) 47
  Planar WRHMR Total
< 14.2% ≥ 14.2%
(C)    
SPECT WRCNN    
 ≤ 30% 18 (38.3%) 1 (2.1%) 19 (40.4%)
 > 30% 12 (25.5%) 16 (34.0%) 28 (59.6%)
Total 30 (63.8%) 17 (36.2%) 47
  1. Planar WRBC, planar washout rate with background correction; Planar WRHMR, planar washout rate heart-to-mediastinum ratio; Planar WRNC, planar washout rate without background correction; SPECT WRCNN, SPECT washout rate using CNN.
Back to article page