UT Southwestern Radiation Oncology...

7/24/2014

1

DEPT OF RADIATION ONCOLOGY

Fast, near real-time, Monte Carlo dose

calculations using GPU

Xun Jia Ph.D.

[email protected]


Outline

GPU Monte Carlo

Clinical Applications

Conclusions

2


Outline

GPU Monte Carlo


Conclusions

3

7/24/2014

2


GPU A specialized accelerator (personal super-computer)

4-16 cores >1000 cores

Comparison CPU/GPU peak processing power

Comparison CPU/GPU performance in linear

algebra solver

4


GPU Advantages

High processing power: 3-4 TFLOPS per card nowadays

Low cost: order of magnitude lower than CPU with similar

processing power

Easy to set up, maintain, and access

Full system control: do not rely on 3rd party management

Low burden of data communication: card is local at the

computer

Suitable for Medical Physics problems

Jia et. al. Review Article, PMB, 59, R151(2014)

5


GPU Monte Carlo project

Since 2009 in UCSD

Utilize GPU to accelerate MC simulations in medical

physics

Use appropriate physics model to maintain accuracy

Design GPU-friendly implementations for high efficiency

Apply developed codes to solve medical physics

problems

6

7/24/2014

3


GPU Monte Carlo project Published >10 papers on peer-reviewed journals, with 6 more

under review/in preparation, and >20 conference presentations

7

Four GPU training courses

Upcoming 5th one in Oct

2014 at UTSW


GPU Monte Carlo project

A wide spectrum of GPU-based MC packages

gDPM: MV photon/electron (2009)

gMCDRR/gCTD: kV photon (2011)

gPMC: proton (+electron) (2012)

gBMC: brachytherapy dose calculations (2014)

g????: MV photon/electron, openCL (2014)

……

Developing new GPU packages

Unifying code interface

Solve clinical problems

8


gDPM gDPM

Maintain DPM physics and hence accuracy

Seek for optimal implementation and high efficiency

Phase space source model with GPU-friendly implementation

Results

Jia et. al., Phys. Med. Biol., 55, 3077 (2010)

Jia et. al., Phys. Med. Biol., 56, 7107 (2011)

Townson et. al., Phys. Med. Biol., 58, 4341(2013)

9

7/24/2014

4


• Average relative uncertainty, t-statistical test passing rate, execution

time T, and speed-up factor

• Multi-GPU

Source type # of Histories Case D/D

(%)

Pall

(%) TCPU (sec) TGPU (sec) TCPU/TGPU

20MeV e 2.5×106 water-lung-water 0.98 99.9 117.5 1.71 69.7

20MeV e 2.5×106 water-bone-water 0.99 99.8 127.0 1.65 77.0

6MV p 2.5×108 water-lung-water 0.72 97.7 1403.7 16.1 87.2

6MV p 2.5×108 water-bone-water 0.64 97.5 1741.0 20.5 84.9

6MV p 2.5×108 VMAT Prostate patient 0.78 N/A N/A 39.6 N/A

6MV p 2.5×108 IMRT HN patient 0.57 N/A N/A 36.1 N/A

Source type # of Histories Case TGPU (sec) T4GPU (sec) TGPU/T4GPU

6MV Photon 4×109 water-lung-water 312.82 78.4 3.99

6MV Photon 4×109 water-bone-water 403.75 101.19 3.99

CPU: Intel Xeon processor with 2.27GHz

GPU: NVIDIA Tesla C2050

10

gDPM


Automatic commissioning

Phase spacelet (PSL)

Adjust PSL weights to match calculation with measurements

11

gDPM Tian et. al., submitted to Phys. Med. Biol. (2014)

0 5 10 15 20 25 30-0.03

0

0.03

0.06

Do

se

diff

ere

nce

(Gy)

Depth (cm)

-0.02

0

0.02

0.04

Do

se

diff

ere

nce

(Gy) 0.1

0.3

0.5

0.7

0.9

1.1

Depth (cm)

Do

se

(G

y)

desired

uncommissioned

commissioned

0 5 10 15 20 25 30

0.2

0.4

0.6

0.8

1

Depth (cm)

Dose (

Gy)

desired

uncommissioned

commissioned0 5 10 15 20 25 30

0.2

0.4

0.6

0.8

1

Depth (cm)

Dose (

Gy)

desired

uncommissioned

commissioned

1.5x1.5

5x5

10x1015x15

1.5x1.5

15x15

(a)

(b)

(c)

-0.03-0.02-0.01

00.010.02

Diff

ere

nce

(Gy)

-12 -6 0 6 120

0.1

0.2

0.3

0.4

0.5

Off-axis distance (cm)

Do

se

(G

y)

4 5 6 7 8 9 10 110

0.2

0.4

0.6

0.8

1


Do

se

(G

y)

desired

uncommissioned

commissioned

4 5 6 7 8 9 10 11-0.04-0.03-0.02-0.01

00.01


Diffe

ren

ce

(Gy)

4 5 6 7 8 9 10 110.04

-0.03-0.02-0.01

00.01


Diffe

ren

ce

(Gy)

-0.04

-0.02

0

0.02

Diffe

ren

ce

(Gy) -0.07

-0.05-0.03-0.010.01

0.03

Diffe

ren

ce

(Gy)

-0.05

-0.03

-0.01

0.01

Diffe

ren

ce

(Gy) -0.06

-0.04

-0.02

0

0.02

Diffe

ren

ce

(Gy) 4 5 6 7 8 9 10 11

0

0.2

0.4

0.6

0.8

1


Do

se

(G

y)

(a1)

(a2)

(a3)

(a4)

(b1)

(b2)

(b3)

(b4)

2.5cm 2.5cm

2.5cm

10cm

10cm 10cm

20cm 20cm

20cm

2.5cm

10cm

20cm

-0.03-0.02-0.01

00.010.02

Diffe

ren

ce

(Gy)

-12 -6 0 6 120

0.1

0.2

0.3

0.4

0.5


Do

se

(G

y)

0

2

4

6

8

10(Gy)

a b

-15 -10 -5 0 5-0.3-0.2-0.1

00.10.2

z position in CT coordinate(cm)

Diffe

ren

ce

(Gy) 0

2

4

6

8

10

Do

se

(Gy)

-15 -10 -5 0 5-0.7

-0.5

-0.3

-0.1

0.1


Diffe

ren

ce

(Gy) 0

2

4

6

8

10

Do

se

(Gy)

desired

uncommissioned

commissioned

0

2

4

6

8

10

Do

se

(Gy)

desired

uncommissioned

commissioned

-15 -10 -5 0 5-0.7

-0.5

-0.3

-0.1

0.1


Diffe

ren

ce

(Gy)

(a1)

(a2)

(b1)

(b2)


gCTD/MCDRR Realistic simulations of CBCT imaging process

Facilitate CBCT related projects

Evaluate patient-specific imaging dose

Primary

Scatter

Noise

CBCT

Time:

1 min per projection simulation

20sec for dose calculations

Jia, et. al., PMB, 57, 577 (2012)

Jia, et. al., Med. Phys., 39, 7368 (2012)

(a) (b)(b)

Ray-tracing MC

Catphan phantom

(a) (b)

(c) (d)

12

7/24/2014

5


gPMC In collaboration with MGH group

Combines the physics from literatures

Time: 6~22 sec of computation time

In the process of clinical evaluations Jia, et. al., Phys. Med. Biol., 57, 7783 (2012)

Jia et. al., AAPM 2013

Giantsoudi, et. al., AAPM 2014

13


Other GPU packages GPUMCD

~2 min dose calc + optimization,

one GPU/beam

GMC:

Geant4 implementation

349 sec on GTX 580

ARCHERRT

60~80 sec on Tesla M2090

14

Hissoiny, et. al., Phys. Med. Biol., 56, 5119 (2011)

Bol et. al., Phys. Med. Biol., 57, 1375 (2012)

Jahnke, et. al., Phys. Med. Biol., 57, 1217 (2012)

Su, et. al., Med. Phys. 41, 071709 (2014)


Outline

GPU Monte Carlo


Conclusions

15

7/24/2014

6


Pencil beam dose calculations using MC

Adaptively sample particles based on optimized MU

Acceleration 4X (in addition to acceleration of MC by GPU)

16

Treatment (re)-planning

Dash – Ground truth: 1×106 / beamlet

Solid - 5×103 / beamlet with 9 iteration

TH-E-BRE-8, Li, et. al.

Optimized fluence map with 1×106 / beamlet

and 1×104 / beamlet/iteration


A web-based service for treatment plan verification

User upload a plan

gDPM runs on a GPU server

Plan dose is verified and a report is generated

17

Pre-treatment


In/post-treatment Real-time simulation of photon dose delivery using (simulated)

online machine log

18

7/24/2014

7


• Accurate evaluation of patient-specific dose in CT scans

(a) (b) (c)

x

y z

(a) (b) (c)

(d) (e) (f)

x

y

z

Montanari, et. al., PMB. 59, 1239 (2014)

(a) (b) (c)

(d) (e) (f)

x

y

z

19

CBCT dose calculations


Estimate scatter signals using MC

2 FDK reconstruction + 1 MC in 30 sec

20

CBCT scatter

0 200 400 600-1200

-1000

-800

-600

-400

-200

0

Pixel Position of the profile

Hu

Valu

e

Xu, et. al., submitted to PMB. (2014)


Outline

GPU Monte Carlo


Conclusions

21

7/24/2014

8


• GPU is a great tool for MC simulations

• Fast, or almost near real-time dose calculation is possible

• Enable new and novel applications

Conclusions

22


Acknowledgement

23

Steve Jiang, Ph.D.

Professor Xun Jia, Ph.D.

Assistant Professor

Zhen Tian, Ph.D.

Instructor Feng Shi, Ph.D.

Postdoctoral fellow

Michael Folkerts

Ph.D. student

Yongbao Li

Ph.D. student Yuan Xu

Ph.D. student

• Research team

• Collaborators

• MGH group

UT Southwestern Radiation Oncology...

Documents

Transcript of UT Southwestern Radiation Oncology...