AimsFileConvert segfault on some DICOM images

[Manik]

Hi,

I just try to open a DICOM image (directly in Anatomist or using AimsFileConvert). In both cases, I get a "segmentation fault".

(gdb) bt
#0 0x42a485d7 in aims::DicomReader<short>::readData () from ***SHFJ_pack-stable-linux-2.3.2/lib/libaimsdicom.so.2.13
#1 0x42a32deb in aims::DicomFormat<short>::read () from ***SHFJ_pack-stable-linux-2.3.2/lib/libaimsdicom.so.2.13
#2 0x40ed7219 in aims::Reader<AimsData<short> >::read () from ***SHFJ_pack-stable-linux-2.3.2/lib/libaimsdata.so.2.13
#3 0x40589f5c in bool loadData<AimsData<short> > () from ***SHFJ_pack-stable-linux-2.3.2/lib/libanatomist.so.1.30
#4 0x40581fa1 in bool loadVolume<short> () from ***SHFJ_pack-stable-linux-2.3.2/lib/libanatomist.so.1.30
#5 0x41015220 in aims::Process::execute () from ***SHFJ_pack-stable-linux-2.3.2/lib/libaimsdata.so.2.13
#6 0x410140d3 in aims::Process::execute () from ***SHFJ_pack-stable-linux-2.3.2/lib/libaimsdata.so.2.13
#7 0x405801f0 in anatomist::ObjectReader::readAims () from ***SHFJ_pack-stable-linux-2.3.2/lib/libanatomist.so.1.30
#8 0x4057c7ef in anatomist::ObjectReader::load () from ***SHFJ_pack-stable-linux-2.3.2/lib/libanatomist.so.1.30
#9 0x40592e00 in anatomist::AObject::load () from ***SHFJ_pack-stable-linux-2.3.2/lib/libanatomist.so.1.30
#10 0x4071d826 in anatomist::Anatomist::loadObject () from ***SHFJ_pack-stable-linux-2.3.2/lib/libanatomist.so.1.30
#11 0x404d7ecc in anatomist::LoadObjectCommand::doit () from ***SHFJ_pack-stable-linux-2.3.2/lib/libanatomist.so.1.30

I can not try with the development version, because it does not open DICOMs (probably a compilation issue).

The other MRI exams (T1 MRI) of the same patient are read without problems.
It is a vascular image from a GE 1.5T MRI, the sequence is called 3D-MOTSA.
I do not know how it was acquired, but I see there is not only one 3D image in this DICOM.
If I open it using another software, I get the expected image on most slices (looks like a T1mri with different gray tones), and on some slices an image which looks like a 3D reconstruction of the head with the vessels highlighted.

I tried to remove those slices from the directory and I tried to load the same sequence from another patient, but Aims still segfaults.

Is there something else I can try to import these images correctly ?

Tank you,

[riviere]

Hi Manik,

Is this image in the usual DICOM way (a file for each slice) or a single file for the whole 3D image ? We have been reported that some Dicom files could be as one 3D file, and we know they cannot be read by our dicom module (yet).
The development version should be able to read Dicom files like the stable one, you have maybe missed a path/environment variable to use your compiled development version.

Can you put your image somewhere (ftp site for instance) so that I can try it ? If you don't have write access to a ftp site, I can indicate one for you.

In the mean time, you can still convert it using another softrware to another format that our tools can read.

Denis

[Guest]

The image has a file for each slice, but there are two types of images in the same directory. About 1/3 of the software I tried to use does separate correctly the 2 types, all other have a problem with it.

By using an INRIA software (a shell script), I converted the DICOM to some Inrimage files, then to GIS (with Yav++).
There seems to be "two exams in one". Those are split in multiple files by the script.

I would like to send you the files, but when I use XMedCon (http://xmedcon.sourceforge.net) to anonymize them, they can be opened with Anatomist... I suppose the software rewrites the DICOM completely and reorders the slices, so Aims can read them...
I can send you the anonymized files on an ftp server if you can give me an access.
But because they work, it may not be very useful. However you can see the "two sequences" if you open this dicom in Anatomist.
I can also come to the SHFJ with the original CD...

Thank you for your answer.

[riviere]

OK, then bring the CD the next time you come to the SHFJ (I guess we'll see you one of these days).
Denis