6/21/2005VG 2005 Volume Graphics: What's in the cards... Michael Meissner (Viatronix)
6/21/2005VG 2005 Volume Graphics: What's in the cards... Michael Meissner (Viatronix)
6/21/2005VG (10?) key questions from 1999 What’s in the cards? In which cards??? Well, it really depends on the perspective In the following: medical applications
6/21/2005VG 2005 Medical Applications Where do physicians use VG? Diagnostic tool (Unable to automatically tell what’s wrong, 3D is ordered in < 5% of all cases because it is time consuming and difficult) Treatment planning tool (what to do?). Verification tool (automatic detection) Interdisciplinary communication tool
6/21/2005VG 2005 Medical Applications Observations: Scientists: explorative visualization (understand problem and find solution) End-users: 1, 2, or 3 clicks to get to the goal (simplicity!!!) (make solution available, robust, reproducible) Strong NEED for automation, consisting of: Segmentation Detection Guidance Visualization (1D, 2D, 3D, 4D?) only fraction of the problem Etc. Where is this (VG) research and what of it is engineering?
6/21/2005VG (10?) key questions from 1999 On the initial questions: Volumes vs. surfaces? Rectilinear, curved, irregular, etc. Definition of reflection, refraction, etc.? (amorphous) Concepts to unify modelling & rendering API Voxel-based radiosity? Feasible? Advantages? Force & touch with volume data? (Deformation?) Volume based digitisation? Frequency domain representations fast enough? Availability of real-time volume rendering? Impact of VG onto other fields?
6/21/2005VG (10?) key questions from 1999 Availability of real-time volume rendering? CPU, GPU, and dedicated hardware able to deal with voxels (up to 512 3? ) This will continue to improve but: CPUs already slow down GPUs will slow down, too Parallel rendering: unavoidable!!! Multiple CPU cores Intel, AMD, IBM (Cell) Will it be SIMD or MIMD? In the future, where is difference of CPU/GPU? Feasability
6/21/2005VG (10?) key questions from 1999 Availability of real-time volume rendering? APIs limited: High-level: Volumizer, TGS Inventor Low-level: OpenGL, DirectX Related libraries limited: Vtk, Itk, ?
6/21/2005VG (10?) key questions from 1999 What would it take to make it verifiable? Reference: Datasets but also what to look for!!! (app domain) Transfer functions View ports Renderings Etc. Framework to build on top/plug into needed!!! (but where should it come from? Who should be gate keeper?) Research community might not capable of this?
6/21/2005VG (10?) key questions from 1999 Impact of VG onto other fields? Who is impacting who? Apps <> VG Is VG is (becoming) application specific VG has become integral part in medicine but only for a fraction of cases because there are simple alternative solutions. A (classical) volume rendered image is not necessarily most meaningful. Simplicity!!!
6/21/2005VG 2005 Impact of VG: radiology Asking radiologists about the most important innovations of the last decade: MIP Thin slab … 3D Primary reason: too many slices to review
6/21/2005VG 2005 Future What is in the cards: Larger data: 16 bit datasets (64 bit OS) Multi-resolution volume rendering Memory management Parallelism Multi-modality scans: Fusion Registration Non Photo-realistic rendering!!! Segmentation User interfaces
6/21/2005VG 2005 If we accomplish to work with the application domain then we might get:
6/21/2005VG (10?) key questions from 1999 What is going on in medical applications? Previously: dedicated high-end workstations Hot (current) trend is client/server: Expensive high-end VG server based (Thin) client machines already available (& cheap) Network is an issue, WWW is a problem Deployment solely dependent on networks