Figure 1.
Examplars of automatically segmented hippocampi rated as a) no or minor defects (defects estimated to be smaller than ∼0.5% of total volume) b) moderate defects (∼0.5–5% of total volume), major defects (greater than 5% of total volume).
Figure 2.
Left and right hippocampal volumes measured by manual tracing (green) and automatic segmentation (yellow).
** significant difference at p<.01.
Figure 3.
Bland-Altman diagrams plotting the difference in hipocampal volumes computed with the manual and automated methods against their average for each hippocampal pairs for the right (top) and left (bottom) hippocampus.
The even scatter suggests there is no systematic error between methodologies beside the difference in absolute volume.
Figure 4.
Randomly selected 3D models of manually traced (green) and automatically segmented (yellow) hippocampus pairs.
The models were produced using the ITK-Snap software package (www.itksnap.org) for the manual tracings and the Slicer package (www.slicer.org) for the automated measures.
Table 1.
Correlations between hippocampal volumes measured by manual tracing and automated segmentation for different levels of segmentation quality.
Table 2.
Multiple regression analyses contrasting association patterns between hippocampal volume (traced or manually segmented) and theoretically relevant individual variables for the whole sample (excluding segmentations with major defects) and for a sub-sample with no or minor segmentation defects.
Table 3.
Multiple regression analyses investigating the relationship between cognitive variables and hippocampal volumes (manual and automated) in samples stratified by sex.
Figure 5.
Manual tracing (left) and automated segmentation (right) of the hippocampus showing boundary differences (1 & 6: CSF inclusions; 2 greater subicular/entorhinal inclusion; 4 & 5 greater partial volume inclusion).