Error in patch2self for single-slice data

[m13slash9]

Description

When calling a patch2self denoising routine for single-slice data, a np.pad fails (presumably in the 2nd dismension)

----> 1 diff_denoised_data = patch2self(diff_data, gradient_table.bvals, model='ols', shift_intensity=True, clip_negative_vals=False, b0_threshold=25)

C:\ProgramData\Anaconda3\envs\dipy_env\lib\site-packages\dipy\denoise\patch2self.py in patch2self(data, bvals, patch_radius, model, b0_threshold, out_dtype, alpha, verbose, b0_denoising, clip_negative_vals, shift_intensity)
    282 
    283     else:
--> 284         train_b0 = _extract_3d_patches(np.pad(data_b0s, ((patch_radius[0],
    285                                               patch_radius[0]),
    286                                               (patch_radius[1],

<__array_function__ internals> in pad(*args, **kwargs)

C:\ProgramData\Anaconda3\envs\dipy_env\lib\site-packages\numpy\lib\arraypad.py in pad(array, pad_width, mode, **kwargs)
    741 
    742     # Broadcast to shape (array.ndim, 2)
--> 743     pad_width = _as_pairs(pad_width, array.ndim, as_index=True)
    744 
    745     if callable(mode):

C:\ProgramData\Anaconda3\envs\dipy_env\lib\site-packages\numpy\lib\arraypad.py in _as_pairs(x, ndim, as_index)
    516     # Converting the array with `tolist` seems to improve performance
    517     # when iterating and indexing the result (see usage in `pad`)
--> 518     return np.broadcast_to(x, (ndim, 2)).tolist()
    519 
    520 

<__array_function__ internals> in broadcast_to(*args, **kwargs)

C:\ProgramData\Anaconda3\envs\dipy_env\lib\site-packages\numpy\lib\stride_tricks.py in broadcast_to(array, shape, subok)
    409            [1, 2, 3]])
    410     """
--> 411     return _broadcast_to(array, shape, subok=subok, readonly=True)
    412 
    413 

C:\ProgramData\Anaconda3\envs\dipy_env\lib\site-packages\numpy\lib\stride_tricks.py in _broadcast_to(array, shape, subok, readonly)
    346                          'negative')
    347     extras = []
--> 348     it = np.nditer(
    349         (array,), flags=['multi_index', 'refs_ok', 'zerosize_ok'] + extras,
    350         op_flags=['readonly'], itershape=shape, order='C')

ValueError: operands could not be broadcast together with remapped shapes [original->remapped]: (4,2)  and requested shape (2,2)

Way to reproduce

For the traceback above the diff_data.shape is (94, 124, 1, 13)

and

gradient_table.bvals.shape is (13,)

then diff_denoised_data = patch2self(diff_data, gradient_table.bvals, model='ols', shift_intensity=True, clip_negative_vals=False, b0_threshold=25) results in the traceback above.

Attempting diff_denoised_data = patch2self(np.concatenate((diff_data,diff_data,diff_data),2), gradient_table.bvals, model='ols', shift_intensity=True, clip_negative_vals=False, b0_threshold=25) works with no errors

Environment

Windows-10-10.0.19043-SP0 Python 3.9.7 | packaged by conda-forge | (default, Sep 29 2021, 19:15:42) [MSC v.1916 64 bit (AMD64)] dipy 1.4.1 NumPy 1.21.3 SciPy 1.7.1 Nibabel 3.2.1 H5py 3.4.0

[skoudoro]

Thank you for pointing that out @m13slash9.

@ShreyasFadnavis will give you feedback ASAP.

[ShreyasFadnavis]

Hi @m13slash9 -- Thanks for pointing this out, I had not thought of this edge case! Your solution : diff_denoised_data = patch2self(np.concatenate((diff_data,diff_data,diff_data),2), gradient_table.bvals, model='ols', shift_intensity=True, clip_negative_vals=False, b0_threshold=25)

should work as a good workaround without hampering the performance!

@skoudoro : Following up with a PR to fix this.

[skoudoro]

Great! Thanks @ShreyasFadnavis !