On 28.05.2012 16:01, Clarkson, Matt wrote:
> Hi there

Hi Matt,


>
> 1. Tag 0020,0037 (patient orientation) is a direction cosines matrix. If
> different slices have slightly different matrices due to rounding error,
> the loaded image is split into multiple volumes:
> https://github.com/MITK/MITK/blob/master/Core/Code/IO/mitkDicomSeriesReader.cpp#L861

Do you have examples (or example data) of how much is "slightly 
different"? I was not aware of such minor differences from a single 
modality so far. But I guess if such images exist, it would be a valid 
correction to accept some minor differences here.

I am not sure how much we should tolerate. We should consider impact on 
distance or volume measurements, which should not be negatively 
influenced by "nice" loading corrections.


>
> 2. In our images, tag 0020,0037 (patient orientation) may be missing.
> https://github.com/MITK/MITK/blob/master/Core/Code/IO/mitkDicomSeriesReader.cpp#L936

This one is harder. The tag is required by the DICOM Image Plane module 
and without it we cannot make sure that we do not mix images of 
different orientations (which do exist, mixed into one series).
Also, the spatial image sorting (shortly after the line you pointed out) 
could not possibly work without that tag. We could guess about the 
correct orientation and expect that our sorting still works somehow, but 
I guess this will surely produce cases that don't work as expected.

My suggestion would be to modify your images, but I'm also curious about 
different solutions.


>
> 3. In our images, tag 0020, 0032 (patient position) may be missing.
> https://github.com/MITK/MITK/blob/master/Core/Code/IO/mitkDicomSeriesReader.cpp#L936

Same as above, just more severe because we could not even sort spatially 
in many cases.

>
> 4. In our images, tag 0020,000d (Series Instance UID) may be missing.
> https://github.com/MattClarkson/MITK/blob/master/Core/Code/IO/mitkDicomSeriesReader.cpp#L853

Also a required tag. Extremely unusual for modalities to produce images 
without that tag. I would not mind much to add a flag/option for the 
loader to ignore that tag and sort only depending on other tags. This 
should be the exception and not the default, however. The simpler 
solution would be to correct your DICOM images. dcmodify can help much here.


>
> 5. Image loads as signed short, but not as unsigned short. i.e. if we
> force the templating to only load as signed short, it looks ok,
> otherwise, the data is not loaded, and image appears blank.
> https://github.com/MattClarkson/MITK/blob/master/Core/Code/IO/mitkDicomSeriesReader.cpp#L152
>

This sounds bad. Do you have example data that you could upload 
somewhere? Of course types should be interpreted correctly.


> 6. Reportedly, image fails to load first time resulting in a blank
> volume, but then simply doing a File Open and trying again loads the
> image correctly.


Also severe. If you can provide example data, I would take a look into it.


> policy. From a code perspective, it might be nice to provide a way to
> easily plug-in a class that decides if we can handle a certain image,
> like a Chain Of Responsibility pattern?

I can imagine something like callback functions at different points to 
influence which images will be loaded into a single mitk::Image or how 
they are spatially sorted or even something that leaves the whole 
loading to a callback. If we find a couple of good places in the loading 
code where to put such modifications safely, we can work that out.


Kind regards
Daniel


-- 
Dr. Daniel Maleike                           Telefon: + 49 6221 647976 3
Mint Medical GmbH, Friedrich-Ebert-Straße 2, 69221 Dossenheim/Heidelberg
Geschäftsführer: Dr. Matthias Baumhauer
Registergericht Mannheim, HRB 709351

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
mitk-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mitk-users

Reply via email to