Merge pull request #51 from Hekstra-Lab/modernize

Add intensity images to tracking

microutil/btrack/btrack.py

@@ -7,7 +7,15 @@
 ]
 
 
-def gogogo_btrack(labels, config_file, radius, tracks_out):
+def gogogo_btrack(
+ labels,
+ config_file,
+ radius,
+ tracks_out=None,
+ intensity_image=None,
+ properties=(),
+ use_weighted_centroid=True,
+):
  """
  Run btrack on a single position timeseries. Write the track data to h5
  files using the build in export capability of btrack.
@@ -22,6 +30,13 @@ def gogogo_btrack(labels, config_file, radius, tracks_out):
  Maximum search radius for btrack
  tracks_out : str
  Path to h5 files where track data will be saved
+ intensity_image: np.ndarray or None
+ Underlying intensity image from which to compute optional features
+ properties : tuple[str,...]
+ Other properties to compute
+ use_weighted_centroid : bool default True
+ Whether to use the weighted centroid for objects if intensity_image is provided.
+ Default True.
 
  Returns
  -------
@@ -30,7 +45,17 @@ def gogogo_btrack(labels, config_file, radius, tracks_out):
  consistently labelled through time
  """
 
- objects = btrack.utils.segmentation_to_objects(labels)
+ objects = btrack.utils.segmentation_to_objects(
+ labels,
+ intensity_image=intensity_image,
+ use_weighted_centroid=use_weighted_centroid,
+ properties=properties,
+ num_workers=4,
+ )
+
+ tracking_updates = ['motion']
+ if len(properties) > 0:
+ tracking_updates.append('visual')
 
  with btrack.BayesianTracker(verbose=False) as tracker:
  tracker.configure_from_file(config_file)
@@ -43,14 +68,16 @@ def gogogo_btrack(labels, config_file, radius, tracks_out):
  tracker.volume = ((0, labels.shape[-2]), (0, labels.shape[-1]), (-1e5, 1e5))
 
  # track them (in interactive mode)
- tracker.track(step_size=100)
+ tracker.track(tracking_updates=tracking_updates, step_size=50)
 
  # generate hypotheses and run the global optimizer
  tracker.optimize()
 
- # tracker.export(tracks_out, obj_type='obj_type_1')
+ if tracks_out is not None:
+ tracker.export(tracks_out, obj_type='obj_type_1')
+
  tracks = tracker.tracks
- # all_tracks = pd.concat([pd.DataFrame(t.to_dict()) for t in tracks])
+
  tracked_labels = btrack.utils.update_segmentation(labels, tracks)
  return tracked_labels
 

microutil/loading/legacy.py

@@ -194,7 +194,7 @@ def dir_to_df(dir):
  if i == 0:
  all_files = df
  else:
- all_files = all_files.append(df)
+ all_files = pd.concat([all_files, df])
 
  else:
  all_files = dir_to_df(data_dir)

microutil/preprocess.py

@@ -95,7 +95,7 @@ def normalize_fluo(imgs, mode_cutoff_ratio=3.0, n_bins=4096, eps=0.01, dims=list
  )
 
 
-def save_dataset(ds, zarr_path, position_slice, scene_char='S'):
+def save_dataset(ds, zarr_path, position_slice, scene_char='S', scene_size=None):
  """
  Dave a dataset into a specific region of an existing xarray zarr store.
  Intended to be used in preprocessing scripts where each scene is being
@@ -114,14 +114,19 @@ def save_dataset(ds, zarr_path, position_slice, scene_char='S'):
  Passed to the region kwarg of xr.Dataset.to_zarr.
  scene_char : str default 'S'
  Name of the scene dimension.
+ scene_size : int or None
+ If the zarr store does not exist but you know the number of scenes from another
+ source, pass it here to create the dataset with the appropriate dimension.
 
  Returns
  -------
  None: just write the dataset to disk.
  """
+ if scene_size is None:
+ existing_ds = xr.open_zarr(zarr_path)
+ scene_size = existing_ds.sizes[scene_char]
 
- existing_ds = xr.open_zarr(zarr_path)
- dummy = _make_dask_dummy(ds).expand_dims({scene_char: existing_ds.sizes[scene_char]})
+ dummy = _make_dask_dummy(ds).expand_dims({scene_char: scene_size})
  _ = dummy.to_zarr(zarr_path, consolidated=True, compute=False, mode='a')
  ds.expand_dims(scene_char).to_zarr(zarr_path, region={scene_char: position_slice}, mode='a')
 

microutil/segmentation.py

@@ -274,7 +274,7 @@ def individualize(
  def _individualize(mask, topology, threshold):
 
  if topology is None:
- topology = -ndi.morphology.distance_transform_edt(mask)
+ topology = -ndi.distance_transform_edt(mask)
 
  peak_idx = peak_local_max(-topology, min_distance, threshold_abs=threshold)
  peak_mask = np.zeros_like(mask, dtype=bool)

microutil/tests/utils.py

@@ -20,4 +20,4 @@ def open_zarr(path):
  Utility to open an xarray dataset from either dir that pytest might be called from.
  If called from root the path will be different than in the test dir
  """
- return xr.open_zarr(str(dir_.joinpath(path)))
+ return xr.open_zarr(str(dir_.joinpath(path)), consolidated=False)