Fig 1.
Illustration of Imiomics compared to a standard analysis approach.
The standard approach analyses only a small amount of the collected imaging information using explicit measurements. Imiomics uses all collected imaging information and allows analyses of relationships to non-imaging data. In addition to more efficient data usage this allows both untargeted and targeted statistical analysis in the whole-body region, i.e. completely new types of imaging studies.
Fig 2.
Flowchart of key steps in the registration process.
Blue: Image data used, Green: constraints used in the registration.
Fig 3.
(a): Absolute fat content, IFAT. (b): Absolute water content IWATER. (c): Body mask IBODY. (d): Bodypart segmentation. (e): Bone segmentation IBONE, bone segments and anatomical landmarks. (f): Body part and bone segment color coding and landmarks used for each bodypart/bone section in the registration procedure.
Fig 4.
Preliminary whole body imaging atlas.
Coronal slices of preliminary whole body imaging atlas—the mean and standard deviation of absolute fat content in male subjects (n = 60) and in female subjects (n = 68), from the POEM study.
Table 1.
Inverse consistency values for MRI scans from the POEM cohort (n = 68 for female and n = 60 for male) for the bodyparts shown in Fig 3. IME is computed on fat content (%) and VME is vector magnitude (mm). The mean and standard deviation IC values of transformations given by and
(see text) are shown.
Table 2.
Dice values obtained by composition of deformation fields for MRI scans from the POEM cohort (n = 68 for female and n = 60 for male) for the bodyparts shown in Fig 3. The fat and water (wat) content images are thresholded at 50% before and after the composed deformation. The Dice coefficient values are computed on these thresholded binary images. The table holds the mean and standard deviation Dice values.
Fig 5.
The anomaly detection was performed by comparing intensities of one male subject with high liver fat content to a preliminary whole body imaging atlas, of subjects with normal liver fat, that holds pointwise distributions of fat content of 50 male subjects. The whole-body imaging atlas is visualized by the mean value and standard deviation.
Fig 6.
The point-by-point P-values (P-maps) of local tissue volume were obtained by two-tailed t-tests between low weight subjects and high weight subjects (20 men and 23 women).
Fig 7.
Coronal slices of maps with point-wise correlation (r-value) between (a) local tissue volume and weight, (b) local tissue volume and total body fat mass measured by bioimpedance analysis (BIA), (c) local tissue volume and lean tissue (weight minus total body fat mass) and (d) fat content and mean liver fat content. These example correlation analyses are based on 60 male and 68 female subjects.
Fig 8.
One coronal slice of the moving, deformed moving and fixed images of absolute fat content is shown together with a difference image between the fixed and deformed moving images. The subject underwent low calorie diet (LCD) and gastric bypass (GBP) between the moving and fixed MRI-scans. The total weight loss was 14 kg.