Converting STPT mosaic datasets from Zarr to TIFF format

Load the module that we are going to use. The class is STPTDataset defined in imaxt_image.datasets

from imaxt_image.datasets import STPTDataset

Define here the location (path) and name (name) of the dataset. Additionaly set the scale that we want to use: 1, 2, 4, 8, 16 or 32 where 1 is the original resolution, 2 is downsampled 2x2 pixels, etc.

path = "/data/meds1_c/storage/processed/stpt/"
name = "20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1"
scale = 8

We read the data as shown below, instantiating the STPTDataset class wih the location of the dataset and the scale (resolution) that we want to read.

ds = STPTDataset(name, path, scale=scale)
ds

<xarray.Dataset>
Dimensions:  (channel: 4, x: 2340, y: 1820, z: 2)
Coordinates:
  * channel  (channel) int64 1 2 3 4
    type     <U6 'mosaic'
  * x        (x) int64 0 1 2 3 4 5 6 7 ... 2333 2334 2335 2336 2337 2338 2339
  * y        (y) int64 0 1 2 3 4 5 6 7 ... 1813 1814 1815 1816 1817 1818 1819
  * z        (z) int64 0 1
Data variables:
    S001     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S002     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S003     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S004     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S005     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S006     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S007     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S008     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S009     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S010     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S011     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S012     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S013     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S014     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S015     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S016     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S017     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S018     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S019     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>
    S020     (z, channel, y, x) float64 dask.array<chunksize=(1, 1, 1040, 1040), meta=np.ndarray>

We can access e.g. the first slide as:

ds["S001"]

<xarray.DataArray 'S001' (z: 2, channel: 4, y: 1820, x: 2340)>
dask.array<astype, shape=(2, 4, 1820, 2340), dtype=uint16, chunksize=(1, 1, 1040, 1040), chunktype=numpy.ndarray>
Coordinates:
  * channel  (channel) int64 1 2 3 4
    type     <U6 'mosaic'
  * x        (x) int64 0 1 2 3 4 5 6 7 ... 2333 2334 2335 2336 2337 2338 2339
  * y        (y) int64 0 1 2 3 4 5 6 7 ... 1813 1814 1815 1816 1817 1818 1819
  * z        (z) int64 0 1

In order to save the slide to a OME.TIFF file use the .to_tiff method:

ds["S001"].to_tiff(name, dir=".")

Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S001_l.8.ome.tiff

where the dir keyword is the directory where the file is to be saved.

In the same way we can define just a region to save:

ds["S001"][5000:7000,6000:9000].to_tiff(f"{name}_cutout")

Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_cutout_S001_l.8.ome.tiff

Finnally combined with a for loop the above can be done for all sections or for a range of sections:

for section in ds:
    section.to_tiff(name, dir=".")

Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S001_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S002_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S003_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S004_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S005_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S006_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S007_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S008_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S009_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S010_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S011_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S012_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S013_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S014_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S015_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S016_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S017_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S018_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S019_l.8.ome.tiff
Written ./20200311_balbc_8020_3dIMC_20x15_2x8umOS_set1_S020_l.8.ome.tiff