dimension-scale-linking-and-cross-group-indexing
Summary
Link HDF5 dataset dimensions to shared index scales across groups to enable standardized cross-group coordinate reference and axis semantics in Cardinal::HDF5 containers. This ensures that feature and pixel axes in intensity matrices are properly annotated and traversable via dimension scales.
When to use
When exporting quantified MSI data as HDF5 containers following the Cardinal::HDF5 layout convention, and you need to establish bidirectional indexing between intensity data (feature-by-pixel matrix) and metadata groups (featureData, pixelData). Specifically, when your intensity dataset has two dimensions that must be semantically linked to external index columns—axis 0 to featureData/feature_id and axis 1 to pixelData/pixel_index—so that downstream tools can navigate from a spectrum to its metadata without hardcoding positional assumptions.
When NOT to use
- Input datasets are already dimension-scaled in a different convention (e.g., already linked to different indices or using custom axis semantics)
- HDF5 file structure does not follow the Cardinal::HDF5 group hierarchy (missing or differently named featureData/pixelData groups)
- Intensity data is sparse or stored in a non-dense rectangular format where 1-based pixel indexing or full Cartesian coordinates are not applicable
Inputs
- intensity dataset (float32, shape n_features × n_pixels) in spectraData group
- featureData group containing feature metadata datasets (feature_id, feature_index, mz, etc.)
- pixelData group containing pixel metadata datasets (pixel_index, x, y, sample_id, etc.)
Outputs
- intensity dataset with attached dimension scales (HDF5 DIMENSION_LIST metadata)
- spectraData group with layout attribute set to 'feature_by_pixel'
- validated cross-group index linkages (feature_id on axis 0, pixel_index on axis 1)
How to apply
After creating the intensity dataset (shape n_features × n_pixels, float32) in the spectraData group, attach dimension scales by: (1) creating or retrieving the index dataset in the target metadata group (e.g., featureData/feature_id as UTF-8 strings, or pixelData/pixel_index as int64); (2) converting the index dataset to a dimension scale with h5py.Dataset.make_scale(name=...) if not already marked; (3) calling h5py.Dataset.dims[axis_number].attach_scale(scale_dataset) on the intensity dataset for each axis, where axis 0 links to featureData/feature_id and axis 1 links to pixelData/pixel_index; (4) adding a layout attribute to intensity specifying 'feature_by_pixel' to document the axis semantics; (5) verifying that dimension scales are properly linked by checking that the HDF5 file's internal DIMENSION_LIST attributes are populated and that tools like h5py can traverse scales without error. The dimension scales must be sorted consistently (e.g., featureData ascending by m/z) and use 1-based indexing conventions where appropriate to align with the Cardinal::HDF5 standard.
Related tools
- LipidQMap (MSI quantitation and HDF5 export engine; orchestrates dimension-scale linking as part of Cardinal::HDF5 container serialization) — https://github.com/swinnenteam/LipidQMap
- Cardinal (Reference standard and downstream consumer of dimension-scaled HDF5 MSI containers; enforces Cardinal::HDF5 layout conventions) — https://cardinalmsi.org
- h5py (HDF5 library API for attaching and verifying dimension scales in Python)
Evaluation signals
- Verify intensity dataset's DIMENSION_LIST attribute is populated with references to featureData/feature_id (axis 0) and pixelData/pixel_index (axis 1) using h5py inspector or h5dump
- Confirm that both featureData/feature_id and pixelData/pixel_index datasets are marked as dimension scales (have CLASS='DIMENSION_SCALE' attribute)
- Check that intensity dataset has a layout attribute with value 'feature_by_pixel'
- Validate that all featureData and pixelData indices align row-wise with intensity axes (n_features and n_pixels match respectively) and are sorted consistently
- Test that a consuming tool (e.g., Cardinal) can traverse the dimension scales and resolve feature/pixel metadata without errors
Limitations
- Dimension scales require that index datasets be created and marked as scales before attachment; if featureData or pixelData groups are missing or incomplete, dimension scale attachment will fail
- 1-based indexing convention must be applied consistently across all pixel coordinates (x, y, pixel_index) and feature indices; mixing 0-based and 1-based indices will break cross-group navigation
- HDF5 dimension scales are a read-only metadata layer; they do not enforce referential integrity or prevent orphaned indices if metadata is later deleted or modified outside the dimension-scale framework
- The Cardinal::HDF5 convention assumes rectangular dense pixel grids; sparse imaging data or irregularly sampled pixels may not map cleanly to standard dimension scales without additional coordinate metadata (min_x, max_x, etc.)
Evidence
- [methods] attach dimension scales linking axis 0 to featureData/feature_id and axis 1 to pixelData/pixel_index: "attach dimension scales linking axis 0 to featureData/feature_id and axis 1 to pixelData/pixel_index, and add group attributes n_features and n_spectra"
- [methods] LipidQMap writes MSI exports as HDF5 containers that follow the Cardinal::HDF5 conventions, enabling standardized storage and interchange of quantified imaging data.: "LipidQMap writes MSI exports as HDF5 containers that follow the Cardinal::HDF5 conventions, enabling standardized storage and interchange of quantified imaging data."
- [methods] write the intensity dataset with feature_by_pixel layout attribute: "write the intensity dataset with feature_by_pixel layout attribute, attach dimension scales linking axis 0 to featureData/feature_id"
- [methods] Validate the output file: verify all required datasets and groups exist, check dataset shapes and dtypes, confirm dimension scales are properly linked: "Validate the output file: verify all required datasets and groups exist, check dataset shapes and dtypes, confirm dimension scales are properly linked, and verify 1-based indexing is used throughout."
1---2name: dimension-scale-linking-and-cross-group-indexing-23description: Use when when exporting quantified MSI data as HDF5 containers following the Cardinal::HDF5 layout convention, and you need to establish bidirectional indexing between intensity data (feature-by-pixel matrix) and metadata groups (featureData, pixelData).4license: CC-BY-4.05---67# dimension-scale-linking-and-cross-group-indexing89## Summary1011Link HDF5 dataset dimensions to shared index scales across groups to enable standardized cross-group coordinate reference and axis semantics in Cardinal::HDF5 containers. This ensures that feature and pixel axes in intensity matrices are properly annotated and traversable via dimension scales.1213## When to use1415When exporting quantified MSI data as HDF5 containers following the Cardinal::HDF5 layout convention, and you need to establish bidirectional indexing between intensity data (feature-by-pixel matrix) and metadata groups (featureData, pixelData). Specifically, when your intensity dataset has two dimensions that must be semantically linked to external index columns—axis 0 to featureData/feature_id and axis 1 to pixelData/pixel_index—so that downstream tools can navigate from a spectrum to its metadata without hardcoding positional assumptions.1617## When NOT to use1819- Input datasets are already dimension-scaled in a different convention (e.g., already linked to different indices or using custom axis semantics)20- HDF5 file structure does not follow the Cardinal::HDF5 group hierarchy (missing or differently named featureData/pixelData groups)21- Intensity data is sparse or stored in a non-dense rectangular format where 1-based pixel indexing or full Cartesian coordinates are not applicable2223## Inputs2425- intensity dataset (float32, shape n_features × n_pixels) in spectraData group26- featureData group containing feature metadata datasets (feature_id, feature_index, mz, etc.)27- pixelData group containing pixel metadata datasets (pixel_index, x, y, sample_id, etc.)2829## Outputs3031- intensity dataset with attached dimension scales (HDF5 DIMENSION_LIST metadata)32- spectraData group with layout attribute set to 'feature_by_pixel'33- validated cross-group index linkages (feature_id on axis 0, pixel_index on axis 1)3435## How to apply3637After creating the intensity dataset (shape n_features × n_pixels, float32) in the spectraData group, attach dimension scales by: (1) creating or retrieving the index dataset in the target metadata group (e.g., featureData/feature_id as UTF-8 strings, or pixelData/pixel_index as int64); (2) converting the index dataset to a dimension scale with `h5py.Dataset.make_scale(name=...)` if not already marked; (3) calling `h5py.Dataset.dims[axis_number].attach_scale(scale_dataset)` on the intensity dataset for each axis, where axis 0 links to featureData/feature_id and axis 1 links to pixelData/pixel_index; (4) adding a layout attribute to intensity specifying 'feature_by_pixel' to document the axis semantics; (5) verifying that dimension scales are properly linked by checking that the HDF5 file's internal DIMENSION_LIST attributes are populated and that tools like h5py can traverse scales without error. The dimension scales must be sorted consistently (e.g., featureData ascending by m/z) and use 1-based indexing conventions where appropriate to align with the Cardinal::HDF5 standard.3839## Related tools4041- **LipidQMap** (MSI quantitation and HDF5 export engine; orchestrates dimension-scale linking as part of Cardinal::HDF5 container serialization) — https://github.com/swinnenteam/LipidQMap42- **Cardinal** (Reference standard and downstream consumer of dimension-scaled HDF5 MSI containers; enforces Cardinal::HDF5 layout conventions) — https://cardinalmsi.org43- **h5py** (HDF5 library API for attaching and verifying dimension scales in Python)4445## Evaluation signals4647- Verify intensity dataset's DIMENSION_LIST attribute is populated with references to featureData/feature_id (axis 0) and pixelData/pixel_index (axis 1) using h5py inspector or h5dump48- Confirm that both featureData/feature_id and pixelData/pixel_index datasets are marked as dimension scales (have CLASS='DIMENSION_SCALE' attribute)49- Check that intensity dataset has a layout attribute with value 'feature_by_pixel'50- Validate that all featureData and pixelData indices align row-wise with intensity axes (n_features and n_pixels match respectively) and are sorted consistently51- Test that a consuming tool (e.g., Cardinal) can traverse the dimension scales and resolve feature/pixel metadata without errors5253## Limitations5455- Dimension scales require that index datasets be created and marked as scales before attachment; if featureData or pixelData groups are missing or incomplete, dimension scale attachment will fail56- 1-based indexing convention must be applied consistently across all pixel coordinates (x, y, pixel_index) and feature indices; mixing 0-based and 1-based indices will break cross-group navigation57- HDF5 dimension scales are a read-only metadata layer; they do not enforce referential integrity or prevent orphaned indices if metadata is later deleted or modified outside the dimension-scale framework58- The Cardinal::HDF5 convention assumes rectangular dense pixel grids; sparse imaging data or irregularly sampled pixels may not map cleanly to standard dimension scales without additional coordinate metadata (min_x, max_x, etc.)5960## Evidence6162- [methods] attach dimension scales linking axis 0 to featureData/feature_id and axis 1 to pixelData/pixel_index: "attach dimension scales linking axis 0 to featureData/feature_id and axis 1 to pixelData/pixel_index, and add group attributes n_features and n_spectra"63- [methods] LipidQMap writes MSI exports as HDF5 containers that follow the Cardinal::HDF5 conventions, enabling standardized storage and interchange of quantified imaging data.: "LipidQMap writes MSI exports as HDF5 containers that follow the Cardinal::HDF5 conventions, enabling standardized storage and interchange of quantified imaging data."64- [methods] write the intensity dataset with feature_by_pixel layout attribute: "write the intensity dataset with feature_by_pixel layout attribute, attach dimension scales linking axis 0 to featureData/feature_id"65- [methods] Validate the output file: verify all required datasets and groups exist, check dataset shapes and dtypes, confirm dimension scales are properly linked: "Validate the output file: verify all required datasets and groups exist, check dataset shapes and dtypes, confirm dimension scales are properly linked, and verify 1-based indexing is used throughout."
HolobiomicsLab