[Insight-developers] Image orientation advice wanted

[Richard Beare]

Hi everyone,

There has been some discussion on this topic lately, so I thought I
had better ask my questions while everyone is warmed to the topic.

I am using a number of image formats combined with a variety of tools
and have concerns that I'm not doing everything I should to ensure
consistency. I'd appreciate any advice people can offer.

The formats I'm currently using are Analyze, nifti and minc. Minc has
been used for a long time in my current group and I'm keen to use the
associated display tools. I'm using AIR for registration (will swap to
ITK as time permits), which reads and writes analyze files and I use
nifti as an intermediate step in translation to and from minc. I use
the nii2mnc and mnc2nii tools distributed with minc libraries for this
translation step.

Many of the templates I'm using are originally in minc format.

My original source images are dicom. I convert these to analyze using
itk as follows:

use GDCMSeriesFileNames to get a list of SeriesUIDs

loop over the list, reading each with a SeriesReader

Write the output with an orienter as follows:
      typedef itk::OrientImageFilter<ImageType,ImageType> OrienterType;
      typename OrienterType::Pointer orienter;
      orienter = OrienterType::New();
      orienter->UseImageDirectionOn();
           orienter->SetDesiredCoordinateOrientation(itk::SpatialOrientation::ITK_COORDINATE_ORIENTATION_RPI);
      orienter->SetInput(imbuff);
      orienter->Update();

Conversion between Analyze and nifti is done using a standard itk read
and write pipeline. Should I use an orienter in this process as well?

I'm concerned that conversion between even this small number of
formats is fraught with danger, so I'd appreciate any advice/warnings
that can be offered.

Thanks