On 24 November 2011 09:34, Andy Buckle <andybuc...@gmail.com> wrote:
> On 23 November 2011 23:05, Kris Thielemans <kris.f.thielem...@gmail.com>
> wrote:
>> Hi
>>
>> I'm modifying octave-forge/extra/dicom/src/dicominfo.cpp to handle data
>> which are arrays, i.e. VM>1. I realize that not many people will know DICOM,
>> let alone GDCM, but I'd appreciate any feedback in any case. (There's some
>> generic octave questions at the end)
>>
>> Also, it would be good if someone could test the current dicominfo on
>> systems with different byteorders. I suspect that it'll only work when the
>> byteorder of the dicom file is the same as the one of the machine you're
>> running it on.
>>
>> I had a look at gdcm example code and it seems to me that we cannot really
>> use memcpy to get from the dicom raw data to octave data. This presumably
>> wouldn't work when byteswapping is needed (or other "transfer syntax"). In
>> any case, I wouldn't know if memcpy would work with arrays (I'm not sure if
>> it's guaranteed by DICOM that the subsequent members are stored
>> consecutively, although it's very likely of course). In an attempt to handle
>> this (and avoid horrible code repetition) I came up with the following
>> modification for element2value() in dicominfo.cpp:
>>
>> ------------------ current code -------------------
>> if (vr & gdcm::VR::US) {// unsigned short
>> uint16_t usval ;
>> memcpy(&usval, elem->GetByteValue()->GetPointer(), 2);
>> *ov=usval;
>> if(chatty) octave_stdout << '[' << usval << "]\n";
>> } else ...
>> // almost repeated for lots of VRs and types
>> return DICOM_OK;
>> -------------------- new code -----------------
>> if (vr & gdcm::VR::US) {// unsigned short
>> return element2intvalueHelper<gdcm::VR::US>(ov, elem, chatty);
>> } else ...
>> // repeated for lots of VRs
>>
>> ---------------------- new function definition (in dicominfo.cpp)
>> ---------------------
>> template <gdcm::VR::VRType vrtype>
>> int element2intvalueHelper(octave_value *ov, const gdcm::DataElement * elem,
>> const int chatty) {
>> if( !elem->IsEmpty() ) {
>> // find type that corresponds to this VR
>> typedef typename gdcm::VRToType<vrtype >::Type actual_type;
>> gdcm::Element<vrtype,gdcm::VM::VM1_n> el;
>> el.Set( elem->GetValue() );
>> intNDArray<octave_int<actual_type> > val(dim_vector (el.GetLength(),1));
>> for (unsigned i=0; i<el.GetLength(); ++i)
>> val(i) = el.GetValue(i);
>> *ov=val;
>> if(chatty) octave_stdout << '[' << val << "]\n";
>> return DICOM_OK;
>> }
>> else
>> {
>> return DICOM_NOTHING_ASSIGNED;
>> }
>> }
>> -------------------------------------------------------------------------
>>
>> This seems to work. It might be somewhat inefficient though, as we're
>> creating a temporary array, copying to an octave_value, and then getting rid
>> of it. I'm not sure how to avoid this, or even if it's an issue.
>> Also, it's annoying that I need to have 2 separate code sets for ints and
>> floats (as I need intNDArray<octave_int<short> > etc because of missing
>> constructors in octave_value).
>>
>> Any suggestions for improvements?
>>
>> Thanks
>>
>> Kris Thielemans
>> Algorithms and Software Consulting Ltd
>> Honorary Lecturer at Imperial College London
>
> It's really nice to see someone else working on the dicom package.
>
> Have you timed it on your system? How much slower is it?
>
> It did occur to me that memcpy may cause problems. Do you have a
> system where it does? I have used it on a few and it always works. I
> guess having someone with different types of mac to have a try might
> be useful.
>
> Do you have write acces to sf.net svn? If not please can this be
> granted. Carnë, is that your remit now?
>
> Driver issues with my linux system are still consuming my spare time,
> so I have not actually tried your code yet. I would not actually
> commit this yet unless;
> - The timed performance degradation is tiny, or,
> - We have confirmation of our suspicion that it does cause a problem
> on some systems.
>
> --
> /* andy buckle */
Hi Kris
do you have any experience with svn? If so, I can give you write
access to the repo as requested by Andy. You'll need a sourceforge
account for that (you can use openID such as your google account to
create one).
Carnë
------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure
contains a definitive record of customers, application performance,
security threats, fraudulent activity, and more. Splunk takes this
data and makes sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-novd2d
_______________________________________________
Octave-dev mailing list
Octave-dev@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/octave-dev
