Skip to main content

Table 1 Overview of the images and reference segmentations used in the different experiments

From: Automatic segmentation of large-scale CT image datasets for detailed body composition analysis

Segmentation target

SCAPIS

IGT

Total

CV

Test

Total

CV

Test

Liver slice

Liver accurateIJ

51

51

Liver crudeAM

2681

2413

268

1951

1756

195

SpleenDP

51

51

Abdomen slice

SATAM

2677

2410

267

1951

1756

195

VATAM

2677

2410

267

1951

1756

195

IPAT

1017

916

101

RPATIJ

1017

916

101

DSATIJ

529

477

52

SSAT

529

477

52

Spine bone marrowDP

208

188

20

Skeleton muscleDP

200

180

20

Thigh slice

SATAM

2682

2414

268

1951

1756

195

MuscleAM

2682

2414

268

1951

1756

195

IMATAM

2682

2414

268

1951

1756

195

Bone marrowa

2683

1951

Cortical bonea

2683

1951

  1. Total total number of images with references segmentation used, CV cross validation, Test test set used—the test split was only used when the total number samples were above 100
  2. AMReference segmentations generated at Antaros Medical, see “Methods” section
  3. IJReference segmentations generated at Uppsala University using the software Image J, see “Methods” section
  4. DPReference segmentations generated at Uppsala University using the software Deep Paint, see “Methods” section
  5. aAutomatic segmentation pipeline developed with traditional image analysis techniques, without the use of deep learning, which is also evaluated separately