python - 如何使用 gdcm 和 Python 解压压缩的 DICOM 图像？

Question

我正在处理压缩的 DICOM 图像，我想在 Ubuntu 14 上使用 Python 2.7 解压。我正在使用 gdcm，我已按照 this link 安装了它。 (sudo apt-get install python-gdcm)

我正在使用this example解压缩图像(至少当我尝试打开它时，ImageJ 将其称为“压缩的 dicom 图像”)，但出现无法解决的错误。代码如下(只是链接中的示例)

import gdcm
import sys

if __name__ == "__main__":
  file1 = sys.argv[1]
  file2 = sys.argv[2]

  r = gdcm.ImageReader()
  r.SetFileName(cin)

  if not r.Read():
    sys.exit(1)

  image = gdcm.Image()
  ir = r.GetImage()

  image.SetNumberOfDimensions( ir.GetNumberOfDimensions() );
  dims = ir.GetDimensions();
  print ir.GetDimension(0);
  print ir.GetDimension(1);
  print "Dims:",dims

  image.SetDimension(0, ir.GetDimension(0) );
  image.SetDimension(1, ir.GetDimension(1) );

  pixeltype = ir.GetPixelFormat();
  image.SetPixelFormat( pixeltype );

  pi = ir.GetPhotometricInterpretation();
  image.SetPhotometricInterpretation( pi );

  pixeldata = gdcm.DataElement( gdcm.Tag(0x7fe0,0x0010) )
  str1 = ir.GetBuffer()
  #print ir.GetBufferLength()
  pixeldata.SetByteValue( str1, gdcm.VL( len(str1) ) )
  image.SetDataElement( pixeldata )

  w = gdcm.ImageWriter()
  w.SetFileName(path_save+"uncompressed.png")
  w.SetFile( r.GetFile() )

  w.SetImage( image )

  if not w.Write():
    sys.exit(1)

在print dims标记程序中确实打印了图像的正确尺寸。但是当它到达 w.SetImage(image) 时，我收到一个错误，并且还收到一堆警告:

Warning: In /build/gdcm-uIgnvq/gdcm-2.6.3/Source/MediaStorageAndFileFormat/gdcmOverlay.cxx, line 205, function void gdcm::Overlay::Update(const gdcm::DataElement&)

Warning: In /build/gdcm-uIgnvq/gdcm-2.6.3/Source/MediaStorageAndFileFormat/gdcmPixmapReader.cxx, line 544, function bool gdcm::DoOverlays(const gdcm::DataSet&, gdcm::Pixmap&)
Bits Allocated are wrong. Correcting.


Error: In /build/gdcm-uIgnvq/gdcm-2.6.3/Source/MediaStorageAndFileFormat/gdcmOverlay.cxx, line 265, function bool gdcm::Overlay::GrabOverlayFromPixelData(const gdcm::DataSet&)
Could not find Pixel Data. Cannot extract Overlay.


Warning: In /build/gdcm-uIgnvq/gdcm-2.6.3/Source/MediaStorageAndFileFormat/gdcmPixmapReader.cxx, line 550, function bool gdcm::DoOverlays(const gdcm::DataSet&, gdcm::Pixmap&)
Could not extract Overlay from Pixel Data


Warning: In /build/gdcm-uIgnvq/gdcm-2.6.3/Source/MediaStorageAndFileFormat/gdcmPixmapReader.cxx, line 575, function bool gdcm::DoOverlays(const gdcm::DataSet&, gdcm::Pixmap&)
Invalid BitPosition: 0 for overlay #0 removing it.


python2.7: /build/gdcm-uIgnvq/gdcm-2.6.3/Source/Common/gdcmObject.h:58: virtual gdcm::Object::~Object(): Assertion `ReferenceCount == 0' failed.

此示例仅对某些类型的图像有效吗？或者我错过了什么？

Answer 1

既然您尝试使用 python 简单地解压缩图像，为什么不简单地使用这个:

import gdcm
import sys

if __name__ == "__main__":
  file1 = sys.argv[1] # input filename
  file2 = sys.argv[2] # output filename

  reader = gdcm.ImageReader()
  reader.SetFileName( file1 )

  if not reader.Read():
    sys.exit(1)

  change = gdcm.ImageChangeTransferSyntax()
  change.SetTransferSyntax( gdcm.TransferSyntax(gdcm.TransferSyntax.ImplicitVRLittleEndian) )
  change.SetInput( reader.GetImage() )
  if not change.Change():
    sys.exit(1)

  writer = gdcm.ImageWriter()
  writer.SetFileName( file2 )
  writer.SetFile( reader.GetFile() )
  writer.SetImage( change.GetOutput() )

  if not writer.Write():
    sys.exit(1)

使用时:

$ python decompress.py gdcm/Testing/Data/012345.002.050.dcm raw.dcm

这导致:

$ gdcminfo raw.dcm 
MediaStorage is 1.2.840.10008.5.1.4.1.1.4 [MR Image Storage]
TransferSyntax is 1.2.840.10008.1.2 [Implicit VR Little Endian: Default Transfer Syntax for DICOM]
NumberOfDimensions: 2
Dimensions: (256,256,1)
SamplesPerPixel    :1
BitsAllocated      :16
BitsStored         :16
HighBit            :15
PixelRepresentation:1
ScalarType found   :INT16
PhotometricInterpretation: MONOCHROME2 
PlanarConfiguration: 0
TransferSyntax: 1.2.840.10008.1.2
Origin: (-85,21.6,108.7)
Spacing: (0.664062,0.664062,1.5)
DirectionCosines: (1,0,0,0,0,-1)
Rescale Intercept/Slope: (0,1)
Orientation Label: CORONAL

<小时/>

更新，看来是原来的bug

gdcmObject.h:58: virtual gdcm::Object::~Object(): Assertion `ReferenceCount == 0' failed.

已在此处解决上游:

• Fix an issue with SmartPointer of an Image in Python