/
Medical Imaging Pipeline Medical Imaging Pipeline

Medical Imaging Pipeline - PowerPoint Presentation

calandra-battersby
calandra-battersby . @calandra-battersby
Follow
351 views
Uploaded On 2018-09-22

Medical Imaging Pipeline - PPT Presentation

YuTing Chen Young kyu Choi Jason Cong and Bingjun Xiao Center for DomainSpecific Computing Center for Future Architectures Research Computer Science Department UCLA ID: 675502

stencil add mul computation add stencil computation mul images amp access reconstruction sqrt registration imaging update pipeline medical pattens

Share:

Link:

Embed:

Download Presentation from below link

Download Presentation The PPT/PDF document "Medical Imaging Pipeline" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.


Presentation Transcript

Slide1

Medical Imaging Pipeline

Yu-Ting Chen, Young-kyu Choi, Jason Cong, and Bingjun XiaoCenter for Domain-Specific ComputingCenter for Future Architectures ResearchComputer Science Department, UCLASlide2

Medical Imaging Pipeline

3D imagesMajor Steps:Imaging reconstructionDenoise/deblurImage registrationImage reconstructionSlide3

Reconstruction

DenoiseRegistrationSegmentationData access patternRandom in EMupdate

Stencil in TVupdateStencilRandom in interp;Stencil in othersStencilMajor computation

FP(

mul

, add)

FP(

mul, add, div, sqrt)FP(mul, add)FP(mul, add, div, sqrt)

Summary of Application Characteristics

3D images

Data access

pattens

Most of the applications contain stencil computation

Some of them has random access

pattens

Floating point computations

Add, multiplication, division,

sqrtSlide4

Image Reconstruction

Forward & Backward Projection

Coordinate & intersection length computation :

Do projection based on the updated images

Update images (voxels)Slide5

Remove noises which under Rician distribution

Solve the minimization problem using gradient decentIterative optimization (nested loop)Stencil computationHardware opt. techniquesTilingData reuse designPipeliningDenoiseSlide6

Registration: fluid registration

Align two images and capture the tumorsFluid PDE equationupdateU / update – stencil; updateV - sweepingSegmentation:Try to find and segment an object of interestgradient decent: stencil computationStencil computation patternsApply similar techniques using in DenoiseRegistration & Segmentation

Dataflow of registrationSlide7

Backup SlidesSlide8

Mapping to FPGA Architecture