search

spatial-image / multiscale-spatial-image

Generate a multiscale, chunked, multi-dimensional spatial image data structure that can serialized to OME-NGFF.
Apache License 2.0
38 stars 8 forks source link
imaging multiscale ngff ome-zarr-converter python scientific-computing xarray

readme

multiscale-spatial-image

Test Notebook tests image image DOI

Generate a multiscale, chunked, multi-dimensional spatial image data structure that can serialized to OME-NGFF.

Each scale is a scientific Python Xarray Dataset, organized into nodes of an Xarray Datatree.

Installation

pip install multiscale_spatial_image

Usage

import numpy as np
from spatial_image import to_spatial_image
from multiscale_spatial_image import to_multiscale
import zarr

# Image pixels
array = np.random.randint(0, 256, size=(128,128), dtype=np.uint8)

image = to_spatial_image(array)
print(image)

An Xarray DataArray. Spatial metadata can also be passed during construction.

<xarray.DataArray 'image' (y: 128, x: 128)> Size: 16kB
array([[170,  79, 215, ...,  31, 151, 150],
       [ 77, 181,   1, ..., 217, 176, 228],
       [193,  91, 240, ..., 132, 152,  41],
       ...,
       [ 50, 140, 231, ...,  80, 236,  28],
       [ 89,  46, 180, ...,  84,  42, 140],
       [ 96, 148, 240, ...,  61,  43, 255]], dtype=uint8)
Coordinates:
  * y        (y) float64 1kB 0.0 1.0 2.0 3.0 4.0 ... 124.0 125.0 126.0 127.0
  * x        (x) float64 1kB 0.0 1.0 2.0 3.0 4.0 ... 124.0 125.0 126.0 127.0
# Create multiscale pyramid, downscaling by a factor of 2, then 4
multiscale = to_multiscale(image, [2, 4])
print(multiscale)

A chunked Dask Array MultiscaleSpatialImage Xarray.

<xarray.DataTree>
Group: /
├── Group: /scale0
│       Dimensions:  (y: 128, x: 128)
│       Coordinates:
│         * y        (y) float64 1kB 0.0 1.0 2.0 3.0 4.0 ... 124.0 125.0 126.0 127.0
│         * x        (x) float64 1kB 0.0 1.0 2.0 3.0 4.0 ... 124.0 125.0 126.0 127.0
│       Data variables:
│           image    (y, x) uint8 16kB dask.array<chunksize=(128, 128), meta=np.ndarray>
├── Group: /scale1
│       Dimensions:  (y: 64, x: 64)
│       Coordinates:
│         * y        (y) float64 512B 0.5 2.5 4.5 6.5 8.5 ... 120.5 122.5 124.5 126.5
│         * x        (x) float64 512B 0.5 2.5 4.5 6.5 8.5 ... 120.5 122.5 124.5 126.5
│       Data variables:
│           image    (y, x) uint8 4kB dask.array<chunksize=(64, 64), meta=np.ndarray>
└── Group: /scale2
        Dimensions:  (y: 16, x: 16)
        Coordinates:
          * y        (y) float64 128B 3.5 11.5 19.5 27.5 35.5 ... 99.5 107.5 115.5 123.5
          * x        (x) float64 128B 3.5 11.5 19.5 27.5 35.5 ... 99.5 107.5 115.5 123.5
        Data variables:
            image    (y, x) uint8 256B dask.array<chunksize=(16, 16), meta=np.ndarray>

Map a function over datasets while skipping nodes that do not contain dimensions

import numpy as np
from spatial_image import to_spatial_image
from multiscale_spatial_image import skip_non_dimension_nodes, to_multiscale

data = np.zeros((2, 200, 200))
dims = ("c", "y", "x")
scale_factors = [2, 2]
image = to_spatial_image(array_like=data, dims=dims)
multiscale = to_multiscale(image, scale_factors=scale_factors)

@skip_non_dimension_nodes
def transpose(ds, *args, **kwargs):
    return ds.transpose(*args, **kwargs)

multiscale = multiscale.map_over_datasets(transpose, "y", "x", "c")
print(multiscale)

A transposed MultiscaleSpatialImage.

<xarray.DataTree>
Group: /
├── Group: /scale0
│       Dimensions:  (c: 2, y: 200, x: 200)
│       Coordinates:
│         * c        (c) int32 8B 0 1
│         * y        (y) float64 2kB 0.0 1.0 2.0 3.0 4.0 ... 196.0 197.0 198.0 199.0
│         * x        (x) float64 2kB 0.0 1.0 2.0 3.0 4.0 ... 196.0 197.0 198.0 199.0
│       Data variables:
│           image    (y, x, c) float64 640kB dask.array<chunksize=(200, 200, 2), meta=np.ndarray>
├── Group: /scale1
│       Dimensions:  (c: 2, y: 100, x: 100)
│       Coordinates:
│         * c        (c) int32 8B 0 1
│         * y        (y) float64 800B 0.5 2.5 4.5 6.5 8.5 ... 192.5 194.5 196.5 198.5
│         * x        (x) float64 800B 0.5 2.5 4.5 6.5 8.5 ... 192.5 194.5 196.5 198.5
│       Data variables:
│           image    (y, x, c) float64 160kB dask.array<chunksize=(100, 100, 2), meta=np.ndarray>
└── Group: /scale2
        Dimensions:  (c: 2, y: 50, x: 50)
        Coordinates:
          * c        (c) int32 8B 0 1
          * y        (y) float64 400B 1.5 5.5 9.5 13.5 17.5 ... 185.5 189.5 193.5 197.5
          * x        (x) float64 400B 1.5 5.5 9.5 13.5 17.5 ... 185.5 189.5 193.5 197.5
        Data variables:
            image    (y, x, c) float64 40kB dask.array<chunksize=(50, 50, 2), meta=np.ndarray>

While the decorator allows you to define your own methods to map over datasets in the DataTree while ignoring those datasets not having dimensions, this library also provides a few convenience methods. For example, the transpose method we saw earlier can also be applied as follows:

multiscale = multiscale.msi.transpose("y", "x", "c")

Other methods implemented this way are reindex, equivalent to the xr.DataArray reindex method and assign_coords, equivalent to xr.Dataset assign_coords method.

Store as an Open Microscopy Environment-Next Generation File Format (OME-NGFF) / netCDF store.

It is highly recommended to use dimension_separator='/' in the construction of the Zarr stores.

store = zarr.storage.DirectoryStore('multiscale.zarr', dimension_separator='/')
multiscale.to_zarr(store)

Note: The API is under development, and it may change until 1.0.0 is released. We mean it :-).

Examples

Development

Contributions are welcome and appreciated.

Get the source code

git clone https://github.com/spatial-image/multiscale-spatial-image
cd multiscale-spatial-image

Install dependencies

First install pixi. Then, install project dependencies:

pixi install -a
pixi run pre-commit-install

Run the test suite

The unit tests:

pixi run -e test test

The notebooks tests:

pixi run test-notebooks

Update test data

To add new or update testing data, such as a new baseline for this block:

dataset_name = "cthead1"
image = input_images[dataset_name]
baseline_name = "2_4/XARRAY_COARSEN"
multiscale = to_multiscale(image, [2, 4], method=Methods.XARRAY_COARSEN)
verify_against_baseline(test_data_dir, dataset_name, baseline_name, multiscale)

Add a store_new_image call in your test block:

dataset_name = "cthead1"
image = input_images[dataset_name]
baseline_name = "2_4/XARRAY_COARSEN"
multiscale = to_multiscale(image, [2, 4], method=Methods.XARRAY_COARSEN)

store_new_image(dataset_name, baseline_name, multiscale)

verify_against_baseline(dataset_name, baseline_name, multiscale)

Run the tests to generate the output. Remove the store_new_image call.

Then, create a tarball of the current testing data

cd test/data
tar cvf ../data.tar *
gzip -9 ../data.tar
python3 -c 'import pooch; print(pooch.file_hash("../data.tar.gz"))'

Update the test_data_sha256 variable in the test/_data.py file. Upload the data to web3.storage. And update the test_data_ipfs_cid Content Identifier (CID) variable, which is available in the web3.storage web page interface.

Submit the patch

We use the standard GitHub flow.

Create a release

This section is relevant only for maintainers.

  1. Pull git's main branch.
  2. pixi install -a
  3. pixi run pre-commit-install
  4. pixi run -e test test
  5. pixi shell
  6. hatch version <new-version>
  7. git add .
  8. git commit -m "ENH: Bump version to <version>"
  9. hatch build
  10. hatch publish
  11. git push upstream main
  12. Create a new tag and Release via the GitHub UI. Auto-generate release notes and add additional notes as needed.