There are some significant changes in our Version 8 if you are upgrading from an older version. Some of the major API changes are in our 3D rendering:
The data is loaded in the GPU when a DicomVolume is created. This is great news for memory management, as only one copy is needed, however many 3D images are made from it, and however many viewers show them. It also means that the underlying DicomImage objects can be released once the volume has been created, freeing up main memory (RAM). There is a potential downside which could occur if the GPU restarts for any reason. We could add an event which will fire in that scenario, to prompt you to reload/remake the 3D volumes. A big beneficial side-effect of this change is that any exception which might be generated when making the 3D texture (e.g. If there is insufficient video memory) will cause standard trappable exceptions from the main code, rather than during rendering.
After the request from some of our developers to have the ability to apply a slab-like thickness to MIP and to add Minimum IP. After some debate and experiments, this has been done by replacing MIP mode by Maximum and Slab by Average (as well as adding Minimum). All 3 modes now use the SlabThickness property if it is set to >0 (If <=0, then it is unused). For Max/Min the SlabThickness defaults to 0 (unused) and for Average, it defaults to 5 (mm).
All rendering uses the same Direct3D device. This is what allows all the changes in (1) above, but it has also forced more rigorous locking and synchronisation of GPU operations, which might remove some random anomalous behaviour seen at times under unusual conditions. Proper shutdown of that single device has proved challenging, so we’d like to hear of any issues regarding application shutdown.
By far the biggest changes in V8 are actually “behind the scenes”, made with the intention of improving the resilience, efficiency and maintainability of DicomObjects, and they include:
- Move to DirectX for the main means of rendering, with fall-back to GDI+ for systems which cannot support DirectX.
- Much closer alignment between the COM & .NET versions - both versions have benefited from the synergy during development.
- Far greater normalisation of the code.
- For 2D images, a 3x2 affine matrix is the primary means to support all display control, including that specified by presentation states, etc. - unifying much of the display pipeline.
- A small change is almost exclusive use of DICOM keywords in place of group and element numbers, to improve code readability.
- All internal Radians are now Degrees for consistency.
General API changes
- Some of the less common Matrix methods and DisplayMatrix have been withdrawn to avoid confusion.
- Many methods which previously took a Matrix as a parameter now take a viewer or a Size instead to reduce ambiguity.
- Standardised Colour palettes have been provided for .NET these are static properties of the DicomObjects.ColourPalettes class, but in COM they are accessed via DicomGlobal.Palette(Name) where Name is HotIron, HotMetalBlue, PetColor or Pet2oStep (or their DICOM well-known UIDs).
- New('…') method in COM for instantiating new objects of the specified type has been discontinued, only the standard (language specific) way of creating objects is now supported.
- Collection’s AddNew() method (e.g. DicomLabels) has been discontinued as a step to remove duplicate functions. Instead, you can initialise and object and add it to its collections object.
- UpdateFullScreen has now been merged into Refresh method
3D (both versions)
- 3D images now have a single class, and hence it is possible to move between the types (MPR, Slab, MIP & VR) by changing the Mode property, without remaking the object. In .NET this is the DicomImage3D class, and in COM it is simply a specialised DicomImage, but created via the DicomVolume.Make3D method.
- One DicomVolume object can be used to make multiple 3D image objects - they share the same memory etc. and are therefore efficient, but have different views - using for 3 plane MPR etc.
- The DX3DType is replaced by RenderMode3D, and the names have been changed to show the increased flexibility:
- MIP is now “Maximum” (and Minimum is also available)
- Slab is now Average
- The DicomVolume constructor no longer takes a decimate parameter (as decimation is not needed/support in the 3D texture GPU mode)
- SetViewPlane has been expanded to have 2 overrides - one is “simple” taking an enumerated plane, an option to re-centre and a relative zoom (this is used for “resetting”) to sagital etc. and a more detailed version which takes a centre point, a normal, an “up” vector and a relative zoom. In both case, zoom is relative to the “auto-zoom” which is calculated to include the whole texture in the view area.
3D (.NET only)
- IDicomImage3D is not needed in its previous form, as the same DicomImage3D is used for all the different 3D image types, but IProjection has been added to hold the transformation methods, which are shared with DicomImageFusion
- The properties and methods specific to 3D which were formally members of the image’s View property have been split into 2 groups:
- Those related to the projection (Zoom, Angle, etc.) which are now part of the Projection property of the image.
- Those specifically related to 3D views (such as Slab thickness) are now part of the PropertiesFor3D property of the image.
- This is totally new to both versions - a fusion image (specific class in .NET - again simply a customised DicomImage in COM) In COM is created from the DicomImages.MakeFusion method.
- This supports the Projection property (COM) or the IProjection interface (.NET) to allow spatial transformations, which are passed on simultaneously to ALL the images in the collection. .NET has the OpacityValues, but in COM all images get equal opacity.