biomechanical simulation using supercomputer for predictive medicine

14
Biomechanical Simulation using Supercomputer for Predictive Medicine Shu TAKAGI Department of Mechanical Engineering & Department of Bioengineering The University of Tokyo K. Sugiyama (RIKEN), S. Ii (Osaka Univ.), K. Okita (Nihon Univ.) , T. Azuma ( Univ. Tokyo), Y. Matsumoto ( Univ. Tokyo )

Upload: celso-furukawa

Post on 20-Jan-2015

195 views

Category:

Education


1 download

DESCRIPTION

Author: Takagi, S. Presentation for the Biomedical & Mechatronics Workshop Forum UTokyo USP, 2013

TRANSCRIPT

Page 1: Biomechanical Simulation using Supercomputer for Predictive Medicine

Biomechanical Simulation using Supercomputer for Predictive Medicine

Shu TAKAGIDepartment of Mechanical Engineering & Department of Bioengineering

The University of Tokyo

K. Sugiyama (RIKEN), S. Ii (Osaka Univ.), K. Okita (Nihon Univ.) , T. Azuma ( Univ. Tokyo), Y. Matsumoto ( Univ. Tokyo )

Page 2: Biomechanical Simulation using Supercomputer for Predictive Medicine

“K (京)” computer in Kobe

10P (=1016) operations/sec= 8×104 (nodes) × 8 (cores) × 8 (operations) × 2×109 (Hz)

multi-level parallelismhttp://jp.fujitsu.com/solutions/hpc/brochures

http://www.fujitsu.com/global/about/tech/k/

10 Peta Flops with 640,000 cores

Page 3: Biomechanical Simulation using Supercomputer for Predictive Medicine

For the Design and Developmentof High Intensity Focused Ultrasound

Therapy (HIFU) Device

Page 4: Biomechanical Simulation using Supercomputer for Predictive Medicine

High Intensity Focused Ultrasound therapy

http://www.prostatecancercentre.co.uk/treatments/hifu.html

•Prostate cancer•Breast cancer

Brain cancer Liver cancer

HIFU therapy has been developed for the treatment of deeply-placed cancer.

Control of the focal point by an array transducer

http://www.imasonic.com/

Displacement of focal point due to the reflection and refraction of ultrasound at the interfaces of bones

Page 5: Biomechanical Simulation using Supercomputer for Predictive Medicine

HIFU simulator

Page 6: Biomechanical Simulation using Supercomputer for Predictive Medicine

Comparison of the focal point

With phase delay

0 2Pressure [MPa]

Pmax=2.2MPa

Without phase delay

0 2Pressure [MPa]

Pmax=1.7MPa

Page 7: Biomechanical Simulation using Supercomputer for Predictive Medicine

Numerical MethodSuitable for Medical Image Data(Full Eulerian FSI simulations)

Gaehtgens et al.(1980) Blood Cells., 6, 799.

Page 8: Biomechanical Simulation using Supercomputer for Predictive Medicine

Fluid-Structure coupling analysis of living body

(Finite-Difference or Finite-Element) Simulation

on Eulerian frame

Diagnostic image (CT, MRI)

Voxel data (Volume Fraction of Constituents)representing multi-component geometry

without Mesh Generation

Background

Page 9: Biomechanical Simulation using Supercomputer for Predictive Medicine

by left Cauchy-Greendeformation tensor

EulerianLagrangianHow is the two-phase distinguished ?

by the boundary of mesh by solid volume fractionHow is the solid deformation described?

by the displacement of material points themselves

・ Sugiyama, Ii et al. (2011) J. Comput . Phys., 230, 596.・ Ii, Sugiyama et al. (2011) Int. J. Numer. Meth. Fluids, 65, 150.

vs.

Full Eulerian approach for Fluid-Structure Interactions

s1

00.5

FluidSolid

FluidSolid

Page 10: Biomechanical Simulation using Supercomputer for Predictive Medicine

The Super-Fast FSI Solver: ZZ-EFSI

ZZ-EFSI achieved the actual speed of 4.5 PETA FLOPS!!!

Software availablle athttp://www.islim.org/islim-dl_e.html

Page 11: Biomechanical Simulation using Supercomputer for Predictive Medicine

Development of MultiscaleThrombosis Simulator

Page 12: Biomechanical Simulation using Supercomputer for Predictive Medicine

Blood flow(continuum mechanics)

Protein(wWF)-Protein (GP1b) binding(stochastic process)

Molecular interaction(molecular dynamics)

Multiscale modeling of initial stage of thromobosis

•Substance diffusion•Metabolic reaction and activation•Morphology change

Page 13: Biomechanical Simulation using Supercomputer for Predictive Medicine

Ht = 5.5 % Ht = 15.3 % Ht = 21.9 %

Adhesion of Platelets in a Shear Flow t 200 (ms) = 800 (s-1)

Page 14: Biomechanical Simulation using Supercomputer for Predictive Medicine

Tokai University: S. Goto, S. Shiozaki, N.TamuraRIKEN: S. Noda, H. Yokota, R. HimenoOsaka University: S. TakeuchiShanghai Jiaotong University: X. Gong, F. LiangPeking University: J. WuYork University: H. Huang

Acknowledgement

References: ・ Sugiyama, el al, J. Comput. Phys., Vol.230 (2010), pp.596-627.・ Takagi et al. J. Appl. Mech. Vol.79 (2011), 010911.・ Ii et al., Int. J. Numerical Method. in Fluids, Vol.65 (2011). pp.43-66.・ Ii et al., Comm. in Cmuput. Phys, Vol.12 (2013), pp.544-576.・ Shiozaki et al. JBSE, Vol.7 (2012), pp.275-283.・ Ii et al., J.Comput Phys, , Vol.231 (2012), pp.2328-2358.