OverflowError: Python integer -1024 out of bounds for uint16

[clead6]

Hi, I am getting this overflow error when converting my image from DICOM to NIFTI. I had quick look through your code, I think one of your conditions for converting to float must be missing when doing the scaling. Let me know if you can fix this.

---

OverflowError Traceback (most recent call last)
Cell In[2], line 4
2 dicom_path = Path("data/practical1/CT-PET-VI-02/CT_for_PET/")
3 output_path = Path("data/practical1/ct_for_pet.nii.gz")
----> 4 dicom2nifti.dicom_series_to_nifti(dicom_path, output_path, reorient_nifti=False)

File c:\Users\clead\anaconda3\envs\dicom2nifti\Lib\site-packages\dicom2nifti\convert_dicom.py:79, in dicom_series_to_nifti(original_dicom_directory, output_file, reorient_nifti)
75 shutil.copytree(original_dicom_directory, dicom_directory)
77 dicom_input = common.read_dicom_directory(dicom_directory)
---> 79 return dicom_array_to_nifti(dicom_input, output_file, reorient_nifti)
81 except AttributeError as exception:
82 raise exception

File c:\Users\clead\anaconda3\envs\dicom2nifti\Lib\site-packages\dicom2nifti\convert_dicom.py:119, in dicom_array_to_nifti(dicom_list, output_file, reorient_nifti)
116 vendor = _get_vendor(dicom_list)
118 if vendor == Vendor.GENERIC:
--> 119 results = convert_generic.dicom_to_nifti(dicom_list, output_file)
120 elif vendor == Vendor.SIEMENS:
121 results = convert_siemens.dicom_to_nifti(dicom_list, output_file)

File c:\Users\clead\anaconda3\envs\dicom2nifti\Lib\site-packages\dicom2nifti\convert_generic.py:274, in dicom_to_nifti(dicom_input, output_file)
270 nii_image, max_slice_increment = _convert_slice_incement_inconsistencies(dicom_input)
271 # do the normal conversion
272 else:
273 # Get data; originally z,y,x, transposed to x,y,z
--> 274 data = common.get_volume_pixeldata(dicom_input)
276 affine, max_slice_increment = common.create_affine(dicom_input)
278 # Convert to nifti

File c:\Users\clead\anaconda3\envs\dicom2nifti\Lib\site-packages\dicom2nifti\common.py:313, in get_volume_pixeldata(sorted_slices)
310 for i, slice_ in enumerate(sorted_slices):
311 # create copy so we don't load all pixel data on the original slice that is kept in memory
312 slice_copy = copy.deepcopy(slice_)
--> 313 slice_data = _get_slice_pixeldata(slice_copy)
314 del slice_copy
315 if combined_dtype is None:

File c:\Users\clead\anaconda3\envs\dicom2nifti\Lib\site-packages\dicom2nifti\common.py:355, in _get_slice_pixeldata(dicom_slice)
353 data[data > max_value] = numpy.bitwise_or(data[data > max_value], invert_value)
354 pass
--> 355 return apply_scaling(data, dicom_slice)

File c:\Users\clead\anaconda3\envs\dicom2nifti\Lib\site-packages\dicom2nifti\common.py:509, in apply_scaling(data, dicom_headers)
500 rescale_intercept = dicom_headers.RescaleIntercept
501 # try:
502 # # this section can sometimes fail due to unknown private fields
503 # if private_scale_slope_tag in dicom_headers:
(...)
507 # except:
508 # pass
--> 509 return do_scaling(data, rescale_slope, rescale_intercept)
510 else:
511 return data

File c:\Users\clead\anaconda3\envs\dicom2nifti\Lib\site-packages\dicom2nifti\common.py:540, in do_scaling(data, rescale_slope, rescale_intercept, private_scale_slope, private_scale_intercept)
537 else:
538 # Determine required range
539 minimum_required, maximum_required = data.min(), data.max()
--> 540 minimum_required = min([minimum_required, minimum_required * rescale_slope + rescale_intercept,
541 maximum_required * rescale_slope + rescale_intercept])
542 maximum_required = max([maximum_required, minimum_required * rescale_slope + rescale_intercept,
543 maximum_required * rescale_slope + rescale_intercept])
545 # Determine required datatype from that

OverflowError: Python integer -1024 out of bounds for uint16

[higumalu]

I attempted to use numpy.astype to convert uint16.
The current attempt is successful. I hope this patch can be included in the package in the future.

dicom2nifti/common.py

    else:
        # Determine required range
        data = data.astype(numpy.int16)
        minimum_required, maximum_required = data.min(), data.max()
        minimum_required = min([minimum_required, minimum_required * rescale_slope + rescale_intercept,
                                maximum_required * rescale_slope + rescale_intercept])
        maximum_required = max([maximum_required, minimum_required * rescale_slope + rescale_intercept,
                                maximum_required * rescale_slope + rescale_intercept])

Another approach is to downgrade the numpy package. I try numpy==1.26.4 can solved this problem