Image Registration Accelerated by the GPU

Activity: Talk or presentation typesLecture and oral contribution

See relations at Aarhus University

Karsten Østergaard Noe - Lecturer

  • Department of Computer Science
Being able to perform a 3D registration of two MRI- or CT-images makes it possible to correlate the positioning of organs in a patient at two different times. This can potentially facilitate an improvement of dose planning procedure in radiotherapy treatment. Unfortunately registration methods capable of handling the large deformations in e.g. the pelvic regions are known to be very time consuming. In this talk it will be demonstrated how modern graphic cards can be used for accelerating the method known as viscous-fluid registration. This will be done by describing the needed computations in a way that exploits the parallelized architecture of these cards. Using a graphics card is very cost-efficient and leads to a significant decrease in computation time.
1 Jun 2006

Event (Conference)

TitleMedical Vision Day

ID: 3566041