Fig 1.
Flow diagram of data selection.
Fig 2.
CT images representing the different methods used.
Red border: edges of originally acquired image. Blue circle: FOV of second image, a follow up CT, acquired for that subject. Dark gray: subcutaneous fat. Medium gray: skeletal muscle.
Fig 3.
Flowchart of datasets and methods used.
Method 2 receives both images and segmentations because it appends an image registration step to Method 1. Method 1 and 2 measures just above the aortic arch, Method 3 at thoracic vertebra 5, 8, and 10.
Fig 4.
Schematic example of extreme difference of two scan FOVs and their overlapping area.
The schematic on the left is an extreme example of shifted FOV between scans, the schematic on the right is a more typical example. Gray: lungs. Red: excluded area. Blue: included area. Yellow: area not in either scan. 1: FOV of first scan. 2: FOV of second scan.
Table 1.
Percentages of scans with a segmentation error in the 50 subjects with the largest errors.
Table 2.
Measurement data of the 562 subjects for the Truncated, Compensated, and Extended FOV methods.
Fig 5.
Bland-Altman plots of difference in area between two scans.
The difference is displayed as a percentage of the mean area of those two scans. The vertical mean line is mean area in cm2 across all scans. The horizontal mean line is the mean difference between two scans.
Table 3.
Mean muscle radiodensities for the 562 subjects as measured by the truncated, compensated and extended FOV methods.
Fig 6.
Radiodensity evaluation between truncated FOV on the left and extended FOV on the right.
The radiodensity is evaluated in the areas marked by the red box in the top right of each method. The truncated FOV method had a mean density in the selected area of −108.79HU and the extended FOV method had a mean density of −108.18HU.