Page MenuHomePhabricator

Check failing dicom data in MITK-Data
Closed, ResolvedPublic


OS: Linux (others not checked)

If we try to load some DICOM data in MITK-Data (e.g. DICOMReader/OT-MONO2-8-colon.dcm), the following error occures.

ERROR: Exception occured when reading file /media/neher/Rumpelkammer/mitk/build/release/MITK-Data/DICOMReader/OT-MONO2-8-colon.dcm:
Invalid tag values when constructing tilt information from origin1 '', origin2 '', and orientation ''

Never the less all current tests pass. Since the tests pass, we should check if:
a) test datasets are valid
b) or why they are unused and if the test must be extended to test the data -> so something in the reader must be fixed (or a proper configuration is missing) to allow the loading of the data set.

Event Timeline

floca renamed this task from Please enter a meaningful title to Check failing dicom data in MITK-Data.Apr 25 2018, 1:03 PM
floca created this task.

The error occures on windows too.

I checked the import of directory DICOMReader/ with Slicer (4.8.1) and i could not load any patient because of an error. But, if just open the file DICOMReader/OT-MONO2-8-colon.dcm as data (not as dicom specifically) it works. So i guess some Dicom-Tags of these files are broken.

kislinsk triaged this task as Normal priority.Nov 22 2018, 11:44 AM

I looked into all DICOM images that make this problems and are located in the DICOMReader/ directory.

All of them are invalid DICOMs compared to the standard. All are missing the tags Image Position (0x0020,0x0032) and Image Orientation (0x0020, 0x0037). These tags are required by the standard. Because they are missing the above mentioned error is thrown.

We have to decide on how to proceed. I see at least the following options:

  1. It is an corrupted file. So we still do not support it. But may be additionally output a log information in the NormalDirectionConsistencySorter (thats where the information is tried to query and just empty strings are returned due to missing tags) that the file is "corrupted and not standard complient due to missing tags (20,32) and (20,37)".
  2. We implement a gracefull option as default where we use default values (zero origin, standard orthonormal orientation).
  3. We provide a aditional reader that can be choosen for corrupted/noncompliant DICOM files that basiclly offers he 2nd option, but not as default.

We have discussed, how to proceed with corrupted images (see above).

  1. Check with an old installer, if our old legacy reader was gracefull or not.
  2. If graceful: we should look how the defaults were defined in the old implementation (take this strategy if it stll make sense) and then go for option 2 (see comment above), but with log warning that one can see that something is rotten.
  3. If not graceful: we go for option 1 (see above) and make an explanatory message to hint that the files are corrupted and why.
kalali added a subscriber: bendinge.
kalali added a subscriber: kalali.

@floca Any progress here? This also happened often with real data from partners and is really annoying because the error message is somewhat unspecific.
I would also like to see this fixed for MITK 2020.

@kalali: As nobody has claimed the task and no new comment emerged, I would say "no".
As the next step is fairly simple:

  1. Check with an old installer, if our old legacy reader was gracefull or not.

Everyone can grab the task and push it to the next level.
If you think it should be 2020, tag it as such.

Update: the root of many cases of this problem are the same as discussed in T27568: US 2D image cannot be read: invalid tilt information. In the mean while some cases (as OT-MONO2-8-colon) work now (due to fixes of T27321 and T27322) others (like CR-MONO1-10-chest-*) still do not work. The reasons and solutions are the same as described in T27568.

Discussion result: We close this task. All known issues still open are covered by T27568: US 2D image cannot be read: invalid tilt information. If needed we can open the task again.

floca claimed this task.