[Dcmlib] Overlay in DICOM images - follow up

[Jean-Pierre Roux]

Emmanuel Olart wrote:

> Hi all,
>
> Following our questions last week on overlays in DICOM images, I 
> received this morning my first image with an overlay in the DICOM header.
> This overlay is a binary mask which is supposed to be a lesion contour.
>
> Here is what I can see using PrinFile:
>
> As it has been guessed last week, this is a secondary capture image :
>
> V 0008|0008[CS] [Image Type] [DERIVED\SECONDARY\MPR\CSA 
> MPR\\CSAPARALLEL\M\ND\NORM]
> S 0008|1140[SQ] [Referenced Image Sequence]
>   |  --- SQItem number 0
>   | V fffe|e000[UL] [Item]
>   | V 0008|1150[UI] [Referenced SOP Class UID] 
> [1.2.840.10008.5.1.4.1.1.4 ]  ==>       [MR Image Storage]
>   | V 0008|1155[UI] [Referenced SOP Instance UID] 
> [1.3.12.2.1107.5.2.30.25641.30000005113007072225000001677]
>
> An icon is included in the header, I don't know if we have yet a 
> simple way to view it ?

We have *not*!
:-(
You may also get icons in a DicomDir (at IMAGE level, or at SERIE level)
No simple way to get the image from an icon.
(you need to get the LUT, also;
Ans sometimes, the icon is *compressed* (not a joke!)
Even for the *image* BuildLUT was moved to PixelReadConvert, therefore 
it's not available for end user.
It's time for us to have a new architecture for gdcm !

> S 0088|0200[SQ] [Icon Image Sequence]
>   |  --- SQItem number 0
>   | V fffe|e000[UL] [Item]
>   | V 0028|0002[US] [Samples per Pixel] [1] x(1)
>   | V 0028|0004[CS] [Photometric Interpretation] [PALETTE COLOR ]
>   | V 0028|0010[US] [Rows] [64] x(40)
>   | V 0028|0011[US] [Columns] [64] x(40)
>   | V 0028|0100[US] [Bits Allocated] [8] x(8)
>   | V 0028|0101[US] [Bits Stored] [8] x(8)
>   | V 0028|0102[US] [High Bit] [7] x(7)
>   | V 0028|0103[US] [Pixel Representation] [0] x(0)
>   | V 0028|1101[US] [Red Palette Color Lookup Table Descriptor] 
> [256\0\8] x(100)
>   | V 0028|1102[US] [Green Palette Color Lookup Table Descriptor] 
> [256\0\8] x(100)
>   | V 0028|1103[US] [Blue Palette Color Lookup Table Descriptor] 
> [256\0\8] x(100)
>   | B 0028|1201[OW] [Red Palette Color Lookup Table Data] 
> [gdcm::Binary data loaded; length = 256]
>   | B 0028|1202[OW] [Green Palette Color Lookup Table Data] 
> [gdcm::Binary data loaded; length = 256]
>   | B 0028|1203[OW] [Blue Palette Color Lookup Table Data] 
> [gdcm::Binary data loaded; length = 256]
>   | B 7fe0|0010[OW] [Pixel Data] [gdcm::NotLoaded Ad.:8894 x(22be) 
> Lgt:4096 x(1000)]
>
> And last : here is all informations for the overlay image :
>
> V 6000|0010[US] [Rows] [484] x(1e4)
> V 6000|0011[US] [Columns] [484] x(1e4)
> V 6000|0015[IS] [Number of Frames in Overlay] [1 ]
> V 6000|0022[LO] [Overlay Description] [Siemens MedCom Object Graphics]
> V 6000|0040[CS] [Overlay Type] [G ]
> V 6000|0050[SS] [Overlay Origin] [1\1] x(1)
> V 6000|0051[US] [Image Frame Origin] [1] x(1)
> V 6000|0100[US] [Overlay Bits Allocated] [1] x(1)
> V 6000|0102[US] [Overlay Bit Position] [0] x(0)
> B 6000|3000[OW] [Overlay Data] [gdcm::NotLoaded Ad.:13122 x(3342) 
> Lgt:29282 x(7262)]
>
> As we can see, this is a binary mask, size is 484x484 (same as the 
> original image).
> Bits allocated = 1 : i.e. one byte is used to store information 
> concerning EIGHT voxels.

--> I don't get what you mean ...
(Any image sample is welcome)
JPRx

>
> Finally the original image data is contained in this file :
>
> B 7fe0|0010[OW] [Pixel Data] [gdcm::NotLoaded Ad.:42416 x(a5b0) 
> Lgt:468512 x(72620)]
>
> If I read this image with gdcm, I of course get only the original 
> image data. I'd like to get also the overlay data, maybe in a separate 
> ImageData (used later to superpose overlay on image).
>
> It's quite easy for me to code something to get the value of the 
> 0x6000, 0x3000 field and to convert it to a char* for VTK display for 
> example. I personnaly think that GDCM should provide a way to get 
> either Icon and Overlay data withj separate methods :
>
> gdcm::FileHelper::GetPixelData()
> gdcm::FileHelper::GetIconData()
> gdcm::FileHelper::GetOverlayData()
>
> What do you think ?
>
> Manu
>
> PS: I'll anonymize one of these images if you want to have a look at it.
>