Apply registration task fails on all non-reference images with `ngio.utils._errors.NgioValueError: Incompatible shapes for patch and slice.`

[jluethi]

An error occurred for Kelvin at the Pelkmans lab when using the new apply registration task (version 2.0.0):

An error occurred.
Original error:
Traceback (most recent call last):
  File "/redacted/fractal-server-env/lib/python3.12/site-packages/fractal_server/runner/executors/call_command_wrapper.py", line 62, in call_command_wrapper
    raise TaskExecutionError(
fractal_server.runner.exceptions.TaskExecutionError: Task failed with returncode=1.
STDERR: 2026-04-21 15:32:57,733; run_fractal_task; INFO; START apply_registration_to_image task
2026-04-21 15:32:57,733; apply_registration_to_image; INFO; Running `apply_registration_to_image` on zarr_url='/redacted/cycle1.zarr/B/05/1', registered_roi_table='registered_well_ROI_table' and reference_acquisition=0. Using overwrite_input=True and register_labels=True
2026-04-21 15:32:58,213; apply_registration_to_image; INFO; Write the registered Zarr image to disk
Traceback (most recent call last):
  File "/redacted/FRACTAL_TASKS_DIR/6/fractal-tasks-core/2.0.0/venv/lib/python3.11/site-packages/fractal_tasks_core/apply_registration_to_image.py", line 311, in <module>
    run_fractal_task(task_function=apply_registration_to_image)
  File "/redacted/FRACTAL_TASKS_DIR/6/fractal-tasks-core/2.0.0/venv/lib/python3.11/site-packages/fractal_task_tools/task_wrapper.py", line 99, in run_fractal_task
    metadata_update = task_function(**pars)
                      ^^^^^^^^^^^^^^^^^^^^^
  File "/redacted/FRACTAL_TASKS_DIR/6/fractal-tasks-core/2.0.0/venv/lib/python3.11/site-packages/pydantic/_internal/_validate_call.py", line 39, in wrapper_function
    return wrapper(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^
  File "/redacted/FRACTAL_TASKS_DIR/6/fractal-tasks-core/2.0.0/venv/lib/python3.11/site-packages/pydantic/_internal/_validate_call.py", line 136, in __call__
    res = self.__pydantic_validator__.validate_python(pydantic_core.ArgsKwargs(args, kwargs))
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/redacted/FRACTAL_TASKS_DIR/6/fractal-tasks-core/2.0.0/venv/lib/python3.11/site-packages/fractal_tasks_core/apply_registration_to_image.py", line 210, in apply_registration_to_image
    new_ome_zarr = _write_registered_ngio_image(acq_ome_zarr, new_zarr_url, roi_pairs)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/redacted/FRACTAL_TASKS_DIR/6/fractal-tasks-core/2.0.0/venv/lib/python3.11/site-packages/fractal_tasks_core/apply_registration_to_image.py", line 75, in _write_registered_ngio_image
    new_image.set_roi(ref_roi, patch)
  File "/redacted/FRACTAL_TASKS_DIR/6/fractal-tasks-core/2.0.0/venv/lib/python3.11/site-packages/ngio/images/_image.py", line 400, in set_roi
    self._set_roi(
  File "/redacted/FRACTAL_TASKS_DIR/6/fractal-tasks-core/2.0.0/venv/lib/python3.11/site-packages/ngio/images/_abstract_image.py", line 500, in _set_roi
    roi_numpy_setter(patch)
  File "/redacted/FRACTAL_TASKS_DIR/6/fractal-tasks-core/2.0.0/venv/lib/python3.11/site-packages/ngio/io_pipes/_io_pipes.py", line 171, in __call__
    return self.set(patch)
           ^^^^^^^^^^^^^^^
  File "/redacted/FRACTAL_TASKS_DIR/6/fractal-tasks-core/2.0.0/venv/lib/python3.11/site-packages/ngio/io_pipes/_io_pipes.py", line 311, in set
    set_slice_as_numpy(
  File "/redacted/FRACTAL_TASKS_DIR/6/fractal-tasks-core/2.0.0/venv/lib/python3.11/site-packages/ngio/io_pipes/_ops_slices.py", line 224, in set_slice_as_numpy
    _check_compatibility_of_shapes(zarr_array.shape, patch.shape, slice_tuple)
  File "/redacted/FRACTAL_TASKS_DIR/6/fractal-tasks-core/2.0.0/venv/lib/python3.11/site-packages/ngio/io_pipes/_ops_slices.py", line 209, in _check_compatibility_of_shapes
    raise NgioValueError(
ngio.utils._errors.NgioValueError: Incompatible shapes for patch and slice.
- Patch shape: (4, 1, 11964, 11993)
- Zarr array shape: (4, 1, 12000, 12000)
- Slice tuple: (slice(None, None, None), slice(0, 1, None), slice(0, 11965, None), slice(8, 12000, None))
- Expected shape: (4, 1, 12000, 12000)[(slice(None, None, None), slice(0, 1, None), slice(0, 11965, None), slice(8, 12000, None))] (4, 1, 11965, 11992)

In an example workflow, it occurred on all the non-reference cycle images, while the reference cycle images were processed successfully.

It looks like something goes wrong with the patch size vs. whole zarr array. We should see if we can reproduce this on a smaller example case @lorenzocerrone

[jluethi]

Registration was performed here on the whole well ROI, not per FOV => very large ROI

[lorenzocerrone]

We had a very similar error recently, I would need to know the roi table valued to debug it. Likely it's a one-off error coming from a legacy non-ngio roi table, is the converter the old cell voyager?

[Kelvin-Groot]

I am using Convert Cellvoyager Multiplexing to OME-Zarr v1.5.5

[jluethi]

Thanks @Kelvin-Groot ! We'll debug a bit on our side and will get back to you once we have a better idea on how we can make the task robust to this issue :)

[lorenzocerrone]

I can confirm that using the new converter the tasks runs correctly. But, I still need some time to investigate a fix that works with the old converter.

[jluethi]

Just to add that this off-by-one ROI loading also affects the illumination correction task when loading data from the old converters. Cheng-Han got a:

ValueError: Error in illumination_correction:
image.shape=(33, 2160, 2561)
flatfield.shape=(2160, 2560)

(now testing with the new converter)