[Insight-users] Reading time series (4D) image from DICOM series

[Martin Kavec]

Hi,

I have modified DicomSeriesImageReadWrite2.cxx from Examples/IO to read all 
the DICOM series in a directory containing a 4D fMRI time series and to print 
sizes of itkImage.

Thanks to great leap in GDCM in the ITK-2.6.0, the time series is nicely 
sorted based on acquisition time. However, itkGDCMImageIO produces only 3D 
itkImage [x; y; z * t] out of time series, while one would expect it to be a 
4D itkImage [x; y; z; t]. The 3rd dimension size is a multiple of a number of 
timepoints despite the itkImage being declared as 

itk::Image< PixelType, Dimension = 4 >.

I think, this issue falls into the itkGDCMImageIO's responsibility. It would 
be nice, if this class could have some heuristics to create correctly sized 
itkImage the same way as itkAnalyzeIO can correctly create 4D itkImages. 
Although I guess it is more complicated to decide on a 4D criteria in case of 
DICOM images.

I uploaded anonymized DICOM 4D [128, 128, 24, 4] series on a web as well as my 
modification of DicomSeriesImageReadWrite2.cxx (+CMakeList.txt), if someone 
would like to have a look at this issue. I would appreciate that.

Thanks.

Martin