PMID- 24841148
OWN - NLM
STAT- MEDLINE
DCOM- 20150903
LR  - 20211021
IS  - 1861-6429 (Electronic)
IS  - 1861-6410 (Linking)
VI  - 10
IP  - 3
DP  - 2015 Mar
TI  - GPU-based multi-volume ray casting within VTK for medical applications.
PG  - 293-300
LID - 10.1007/s11548-014-1069-x [doi]
AB  - PURPOSE: Multi-volume visualization is important for displaying relevant 
      information in multimodal or multitemporal medical imaging studies. The main 
      objective with the current study was to develop an efficient GPU-based 
      multi-volume ray caster (MVRC) and validate the proposed visualization system in 
      the context of image-guided surgical navigation. METHODS: Ray casting can produce 
      high-quality 2D images from 3D volume data but the method is computationally 
      demanding, especially when multiple volumes are involved, so a parallel GPU 
      version has been implemented. In the proposed MVRC, imaginary rays are sent 
      through the volumes (one ray for each pixel in the view), and at equal and short 
      intervals along the rays, samples are collected from each volume. Samples from 
      all the volumes are composited using front to back alpha-blending. Since all the rays 
      can be processed simultaneously, the MVRC was implemented in parallel on the GPU 
      to achieve acceptable interactive frame rates. The method is fully integrated 
      within the visualization toolkit (VTK) pipeline with the ability to apply 
      different operations (e.g., transformations, clipping, and cropping) on each 
      volume separately. The implemented method is cross-platform (Windows, Linux and 
      Mac OSX) and runs on different graphics card (NVidia and AMD). The speed of the 
      MVRC was tested with one to five volumes of varying sizes: 128(3), 256(3), and 
      512(3). A Tesla C2070 GPU was used, and the output image size was 600 x 600 
      pixels. The original VTK single-volume ray caster and the MVRC were compared when 
      rendering only one volume. RESULTS: The multi-volume rendering system achieved an 
      interactive frame rate (> 15 fps) when rendering five small volumes (128 (3) 
      voxels), four medium-sized volumes (256(3) voxels), and two large volumes (512(3) 
      voxels). When rendering single volumes, the frame rate of the MVRC was comparable 
      to the original VTK ray caster for small and medium-sized datasets but was 
      approximately 3 frames per second slower for large datasets. The MVRC was 
      successfully integrated in an existing surgical navigation system and was shown 
      to be clinically useful during an ultrasound-guided neurosurgical tumor 
      resection. CONCLUSIONS: A GPU-based MVRC for VTK is a useful tool in medical 
      visualization. The proposed multi-volume GPU-based ray caster for VTK provided 
      high-quality images at reasonable frame rates. The MVRC was effective when used 
      in a neurosurgical navigation application.
FAU - Bozorgi, Mohammadmehdi
AU  - Bozorgi M
AD  - Department of Computer and Information Science, Norwegian University of Science 
      and Technology, Sem Saelandsvei 7-9, 7491 , Trondheim, Norway, 
      mohammeb@idi.ntnu.no.
FAU - Lindseth, Frank
AU  - Lindseth F
LA  - eng
PT  - Journal Article
DEP - 20140520
PL  - Germany
TA  - Int J Comput Assist Radiol Surg
JT  - International journal of computer assisted radiology and surgery
JID - 101499225
SB  - IM
MH  - *Algorithms
MH  - *Data Display
MH  - Equipment Design
MH  - Humans
MH  - Imaging, Three-Dimensional/*methods
MH  - *Multimodal Imaging
MH  - Surgery, Computer-Assisted/*instrumentation
EDAT- 2014/05/21 06:00
MHDA- 2015/09/04 06:00
CRDT- 2014/05/21 06:00
PHST- 2014/01/10 00:00 [received]
PHST- 2014/04/28 00:00 [accepted]
PHST- 2014/05/21 06:00 [entrez]
PHST- 2014/05/21 06:00 [pubmed]
PHST- 2015/09/04 06:00 [medline]
AID - 10.1007/s11548-014-1069-x [doi]
PST - ppublish
SO  - Int J Comput Assist Radiol Surg. 2015 Mar;10(3):293-300. doi: 
      10.1007/s11548-014-1069-x. Epub 2014 May 20.