Value Error while reading dicom_seg-DICOMSEG.dcm

[Suchi97]

Hi Team,

We used the MONAI Bundle to deploy the model using the MONAI Deploy App SDK. As the output, we obtained the dicom_seg-DICOMSEG.dcm. We would like to visualize this output using the pydicom or anyother supported libraries.

We tried to visualize the dcm file as following

import matplotlib.pyplot as plt
import pydicom
 
dataset = pydicom.dcmread('output/dicom_seg-DICOMSEG.dcm')

print("Storage type.....:", dataset.SOPClassUID)
print()
 
if 'PixelData' in dataset:
    rows = int(dataset.Rows)
    cols = int(dataset.Columns)
    print("Image size.......: {rows:d} x {cols:d}, {size:d} bytes".format(
        rows=rows, cols=cols, size=len(dataset.PixelData)))
    if 'PixelSpacing' in dataset:
        print("Pixel spacing....:", dataset.PixelSpacing)

plt.imshow(dataset.pixel_array, cmap=plt.cm.gray)
plt.show()

But we obtained the following error

Storage type.....: 1.2.840.10008.5.1.4.1.1.66.4

Image size.......: 440 x 440, 518616 bytes
Pixel spacing....: [1, 1]
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Input In [1], in <cell line: 17>()
     14     if 'PixelSpacing' in dataset:
     15         print("Pixel spacing....:", dataset.PixelSpacing)
---> 17 plt.imshow(dataset.pixel_array, cmap=plt.cm.gray)
     18 plt.show()

File ~/miniconda3/lib/python3.9/site-packages/pydicom/dataset.py:1882, in Dataset.pixel_array(self)
   1867 @property
   1868 def pixel_array(self) -> "numpy.ndarray":
   1869     """Return the pixel data as a :class:`numpy.ndarray`.
   1870 
   1871     .. versionchanged:: 1.4
   (...)
   1880         :class:`numpy.ndarray`.
   1881     """
-> 1882     self.convert_pixel_data()
   1883     return cast("numpy.ndarray", self._pixel_array)

File ~/miniconda3/lib/python3.9/site-packages/pydicom/dataset.py:1444, in Dataset.convert_pixel_data(self, handler_name)
   1442     self._convert_pixel_data_using_handler(handler_name)
   1443 else:
-> 1444     self._convert_pixel_data_without_handler()

File ~/miniconda3/lib/python3.9/site-packages/pydicom/dataset.py:1556, in Dataset._convert_pixel_data_without_handler(self)
   1547 self._pixel_id = {}
   1549 logger.info(
   1550     "Unable to decode the pixel data using the following handlers: {}."
   1551     "Please see the list of supported Transfer Syntaxes in the "
   (...)
   1554     .format(", ".join([str(hh) for hh in available_handlers]))
   1555 )
-> 1556 raise last_exception

File ~/miniconda3/lib/python3.9/site-packages/pydicom/dataset.py:1536, in Dataset._convert_pixel_data_without_handler(self)
   1534 for handler in available_handlers:
   1535     try:
-> 1536         self._do_pixel_data_conversion(handler)
   1537         return
   1538     except Exception as exc:

File ~/miniconda3/lib/python3.9/site-packages/pydicom/dataset.py:1563, in Dataset._do_pixel_data_conversion(self, handler)
   1559 """Do the actual data conversion using the given handler."""
   1561 # Use the handler to get a 1D numpy array of the pixel data
   1562 # Will raise an exception if no pixel data element
-> 1563 arr = handler.get_pixeldata(self)
   1564 self._pixel_array = reshape_pixel_array(self, arr)
   1566 # Some handler/transfer syntax combinations may need to
   1567 #   convert the color space from YCbCr to RGB

File ~/miniconda3/lib/python3.9/site-packages/pydicom/pixel_data_handlers/numpy_handler.py:330, in get_pixeldata(ds, read_only)
    326         warnings.warn(
    327             "The odd length pixel data is missing a trailing padding byte"
    328         )
    329     else:
--> 330         raise ValueError(
    331             "The length of the pixel data in the dataset ({} bytes) "
    332             "doesn't match the expected length ({} bytes). "
    333             "The dataset may be corrupted or there may be an issue "
    334             "with the pixel data handler."
    335             .format(actual_length, padded_expected_len)
    336         )
    337 elif actual_length > padded_expected_len:
    338     # PS 3.5, Section 8.1.1
    339     msg = (
    340         "The length of the pixel data in the dataset ({} bytes) indicates "
    341         "it contains excess padding. {} bytes will be removed from the "
    342         "end of the data"
    343         .format(actual_length, actual_length - expected_len)
    344     )

ValueError: The length of the pixel data in the dataset (518616 bytes) doesn't match the expected length (1161600 bytes). The dataset may be corrupted or there may be an issue with the pixel data handler.

Please take a look at this issue.

Any help is appreciated.
Thanks in advance.

[Suchi97]

Hi Team,

We tried to load the dicom_seg-DICOMSEG.dcm using MONAI Image reader (ITK Reader)
And we got multiple spleen in one slice . One of the slices of the dicom image is shown below.

We also noticed that the input image shape has 515 slices. And the output segmentation image has only 48 slices.
Please help us to understand this issue.

Any help is appreciated.
Thanks in advance

[MMelQin]

@Suchi97 Thanks for the additional info.

• DICOM pixel data is limited to 16 bytes long, and the metadata specifies the allocated and used bytes. The DICOM Seg object out of the App SDK Seg writer allocates and uses all 16 bytes.
• I was actually about to suggest using (Simple) ITK to read DICOM image and then convert to NIfTI or MHD for further processing, as its GDCM reader can deal with the metadata and pixel data correctly (for most object type)
• The version of the App SDK Seg writer that you use may be the version that had issues, which was addressed in release 0.3
• Better yet, a new version of the App SDK Seg Writer is available in the main branch. Before it is released with 0.5 PyPi package, you can try to download and use the code as a custom operator in your app. The output has been verified with various DICOM viewers as well as DICOM validation tool.
• Example apps in the main branch are also updated to make use of the updated DICOM Seg Writer, which only need minimal change in passing an object to describe the segments instead of just name only.

[dbericat]

@Suchi97 please let us know if you have any more questions or you are good with @MMelQin 's pointers. Thanks!