Ngff table export tracks

README.md

@@ -46,3 +46,21 @@ Add Notebook Link
 
 To add a link to the images with ROIs, run [add_notebook_link.py](scripts/add_notebook_link.py)
 specifying the project ID as a parameter.
+
+
+Export tracks to OME-NGFF
+-------------------------
+
+To export a sample image as an OME-NGFF with tracks stored in AnnData tables
+for viewing in `napari`. First, use `omero-cli-zarr` plugin to export the Image:
+
+    $ omero zarr export Image:13425213
+
+We need `ome-ngff-tables-prototype` and it is safest to pin to a known commit:
+
+    $ pip install git+https://github.com/kevinyamauchi/ome-ngff-tables-prototype.git@51cc60e22807e67253ffcbebbe6dabb597fb94dd
+    $ python path/to/notebooks/idr0113_to_ngff_table.py
+
+Then open in napari:
+
+    $ napari --plugin napari-ome-zarr 13425213.zarr

notebooks/idr0113_to_ngff_table.py

@@ -0,0 +1,134 @@
+
+import requests
+import pandas as pd
+import numpy as np
+import anndata as ad
+from scipy.sparse import csr_matrix
+from ome_zarr.io import parse_url
+import zarr
+from ngff_tables_prototype.writer import write_table_points
+
+from omero.cli import cli_login
+from omero.gateway import BlitzGateway
+
+imageId = 13425213
+
+ROI_URL = "https://idr.openmicroscopy.org/api/v0/m/images/{key}/rois/?limit=500"
+TABLE_URL = "https://idr.openmicroscopy.org/webgateway/table/Image/{key}/query/?query=*"
+
+
+def process_image(conn, imageId):
+    image = conn.getObject("Image", imageId)
+    pixels_id = image.getPrimaryPixels().getId()
+    query = "from PlaneInfo as Info where pixels.id='" + str(pixels_id) + "'"
+    infos = conn.getQueryService().findAllByQuery(query, None)
+
+    times = {}
+    for i in infos:
+        times.update({i.theT.getValue(): int(i.deltaT.getValue())})
+
+    qs = {'key': imageId}
+    url = ROI_URL.format(**qs)
+    json_data = requests.get(url).json()
+
+    # Look-up shapes from their ROI ID
+    roi_map = {}
+    # parse the json
+    for d in json_data['data']:
+        roi_id = d['@id']
+        for s in d["shapes"]:
+            # Assume only ONE shape per ROI
+            roi_map.update({roi_id: s})
+
+    url = TABLE_URL.format(**qs)
+    json_data = requests.get(url).json()
+
+    df = pd.DataFrame(columns=json_data['data']['columns'])
+    for r in json_data['data']['rows']:
+        df.loc[len(df)] = r
+
+    roi_ids = df["Roi"].tolist()
+
+    # Use "Roi" column as index...
+    col = df.loc[:, 'Roi']
+    col.transpose()
+
+    # Add ROIs (shapes) to X table and obs table
+    X_df = pd.DataFrame({c: pd.Series(dtype='float') for c in ["t", "z", "y", "x"]}, index = col.astype(str))
+
+    obs_columns = ["Cell Type", "Cell Uncertainty", "Mother Uncertainty", "Sister Uncertainty", "Roi Name"]
+    obs_df = pd.DataFrame(columns=obs_columns, index = col.astype(str))
+
+
+    # create sparse data...
+    row_count = len(roi_ids)
+    sparse_grid = np.zeros((row_count, row_count), dtype=np.int8)
+
+
+    for index, row in df.iterrows():
+        id = str(row["Roi"])
+        shape = roi_map[row["Roi"]]
+
+        # get coordinates for X
+        dims = ["TheT", "TheZ", "Y", "X"]
+        values = [shape.get(dim) for dim in dims]
+        # roi id as index
+        X_df.loc[id] = values
+
+        # other columns for obs
+        obs_values = [row[col] for col in obs_columns]
+        obs_df.loc[id] = obs_values
+
+        # lineage between points - save to sparse_grid
+        lineage = row["Lineage Roi"]
+        mother = row["Mother Roi"]
+        to_idx = roi_ids.index(row["Roi"])
+        from_id = None
+        if lineage != "":
+            from_id = int(lineage)
+        elif mother != "":
+            from_id = int(mother)
+
+        if from_id is not None:
+            from_idx = roi_ids.index(from_id)
+            sparse_grid[from_idx, to_idx] = 1
+
+
+    print(X_df)
+
+    print(obs_df)
+
+    print(sparse_grid)
+
+    adata = ad.AnnData(X = X_df, obs = obs_df)
+
+    # write as sparse data to obsp
+    adata.obsp["tracking"] = csr_matrix(sparse_grid)
+
+    store = parse_url(str(imageId) + ".zarr", mode="w").store
+    root = zarr.group(store=store)
+
+    tables_group = root.create_group(name="tables")
+    write_table_points(
+        group=tables_group,
+        adata=adata
+    )
+
+
+def main(args):
+    # parser = argparse.ArgumentParser()
+    # parser.add_argument('image', type=int)
+    # args = parser.parse_args(args)
+    # dataset_id = args.dataset
+
+    with cli_login() as cli:
+        conn = BlitzGateway(client_obj=cli._client)
+
+        # image = conn.getObjects('Image', opts={'dataset': dataset_id})
+
+        process_image(conn, imageId)
+
+
+if __name__ == '__main__':
+    import sys
+    main(sys.argv[1:])