Image preparation
Image preparation converts UK Biobank CMR source files into subject-level, analysis-ready files with stable modality names. It is the first contract between UKB source fields and downstream segmentation/feature extraction.
- Modality
- All modalities
- Pipeline step
- Step 1 preparation
- Outputs
- Prepared NIfTI images, optional retained DICOM folders, visit-specific subject directories, Slurm preparation scripts
- Maturity
- Source-audited method page
Current implementation boundary
step1_prepare_data_cmr.py generates Slurm scripts for visit 1 and, when configured, retest/visit 2. It maps raw modality groups to UKB CMR fields: scout 20207, long-axis 20208, short-axis 20209, aortic distensibility 20210, tagging 20211, LVOT 20212, blood flow 20213, and ShMOLLI/native T1 20214.
The generated scripts call scripts/prepare_data.py with modality-specific input folders and options such as --overwrite and --keepdicom. Downstream steps assume the prepared subject directories contain stable filenames such as sa.nii.gz, la_4ch.nii.gz, aortic_flow.nii.gz, lvot_processed.nii.gz, and shmolli_t1map.nii.gz where applicable.
Contract with downstream steps
Prepared images must preserve frame order, temporal spacing, voxel geometry, and modality naming. Segmentation and feature extraction scripts do not rediscover arbitrary filenames; they look for expected files inside the subject directory.
Geometry preservation
Silent affine, spacing, direction, origin, or frame-order changes can invalidate segmentation and measurement outputs. Geometry changes must be explicit and documented before downstream phenotype interpretation.
Prepared modality groups
| Source group | UKB field | Typical downstream use |
|---|---|---|
| Scout/localizer | 20207 | Aortic structure and vascular geometry |
| Cine long-axis | 20208 | Atrial structure/function, long-axis strain, annular and cross-chamber rows |
| Cine short-axis | 20209 | Ventricular volume, mass, function, wall thickness, SAX mechanics |
| Aortic distensibility cine | 20210 | Aortic area, equivalent diameter, pressure-normalized distensibility |
| Tagged MRI | 20211 | Tagged deformation and timing rows |
| LVOT cine | 20212 | Aortic annulus, sinuses, sinotubular junction |
| Phase-contrast flow | 20213 | Aortic velocity, flow, valve area, regurgitation, flow displacement |
| Native T1 mapping | 20214 | Native myocardial and blood-pool T1 |
Source audit
- Raw field mapping and script-generation behavior were checked against
step1_prepare_data_cmr.py. - Downstream filename expectations were checked against current modality pages and feature extraction sources.
- Textbook context boundary: broad clinical textbook context is not surfaced here because this is a pipeline preparation method page.