[Insight-developers] ImageFileReader Bug 0007520 and some considerations/ideas about ImageSeriesReader/Writer

[Mathieu Malaterre]

On Mon, Nov 17, 2008 at 3:34 PM, Iván Macía <imacia at vicomtech.org> wrote:
> Dear all,
>
>
>
> We are trying to read slice by slice a DICOM series. We cannot use
> ImageSeriesReader because the series is heterogeneus in size and it does not
> correspond to a 3D volume. When trying to read the last image with
> ImageFileReader an exception occurs in
> ImageFileReader::EnlargeOutputRequestedRegion()
>
>
>
> "ImageIO returns IO region that does not fully contain the requested region"
>
>
>
>
> Having a look at the source of the problem, ImageFileReader fails when
> trying to read a second image that is smaller that the previous image read.
> The problem is that the LargestPossibleRegion is assigned the new smaller
> size but the RequestedRegion is not. Having a look at Mantis I found a
> related bug
>
>
>
> Bug 0007520: failed propagation of largestPossibleRegion to
> requestedPossibleRegion
>
>
>
> Probably this can be solved with a call to
> output->SetRequestedRegionToLargestPossibleRegion() at the end of
> ImageFileReader::GenerateOutputInformation() after setting
> output->SetLargestPossibleRegion().
>
>
>
> On the other hand, I wanted to know if there are some plans to extend the
> capabilities of ImageSeriesReader and ImageSeriesWriter. We are developing a
> viewer for a DICOM PACS and found many different types of series that does
> not correspond to the usual ITK concept of a series consisting of slices of
> a 3D images. For example, we have found:
>
>
>
> -          Series of 2D images as a time series (cine) or single images with
> multiple frames (these all can be considered as multi-frame).
>
> -          Series with intermixed volumes, such as those that show different
> times of a contrast-enhanced volume. These could also be considered as 3D+t.
>
> -          Series that represent a reslicing of the original volume and show
> a preview image.
>
>
>
> and even more exotic series such as those consisting of
>
>
>
> -          Intermixing of multi-frame and non-multiframe 2D images in the
> same series.
>
> -          Series with different pixel types (for example the preview image
> of the reconstruction can be RGB whereas the rest is short for CT or ushort
> for MR).
>
> -          Images of type RGB<short>.
>
>
>
> I have found some methods in GDCM 1.x that can deal with some of these
> features, such as gdcm::SerieHelper::SplitOnOrientation() and we had to
> develop some of heuristics by our own, but wanted to know if there were some
> plans to add functionality for this analysis in ITK. One idea could be to
> develop a class that could take the file names of a series and analyse them
> in order to split them in groups according to some property.

You have used itk::GDCMSeriesFilenames, right ? If this is not working
for you, please send the full src code + demo data sample + expected
output. As you may have realize this is a non-trivial task to come up
with a general solution, as multiframes DICOM is totally different
from isotropic volume in ITK.

Thanks,
-- 
Mathieu