Mobile Wireless Ultrasound with GPU Beamforming Jesper Mosegaard , - PowerPoint PPT Presentation

Click to edit Master title style Mobile Wireless Ultrasound with GPU Beamforming Jesper Mosegaard , PhD Head of Computer Graphics Lab Alexandra Institute Denmark

The Alexandra Institute Click to edit Master title style • Private not-for-profit company within IT – Technology transfer from University research through GTS institutes (Danish model) • Application oriented research • Consultancy for companies

FutureSonic Click to edit Master title style • ” A new platform and business model for on-demand diagnostic ultrasound imaging“ • 2013-2018 • Budget: 23 mio US$ 19/03/15 3

Joint work Click to edit Master title style • Center for Fast Ultrasound Imaging, Technical University of Denmark – Martin Christian Hemmsen – Borislav G. Tomov – Jørgen Arendt Jensen • BK Medical – Carsten Kjær • Computer Graphics Lab, Alexandra Institute – Thomas Kim Kjeldsen – Lee Lassen – Jesper Mosegaard 19/03/15 Page 4

Presentation based on publications Click to edit Master title style T. Kjeldsen, L. Lassen, M. C. Hemmsen, C. Kjaer, B. G. Tomov, J. Mosegaard, and J. A. Jensen, “Synthetic aperture sequential beamforming implemented on multi-core platforms,” in Proceedings of 2014 ieee international ultrasonics symposium , 2014, pp. 2181-2184. M. C. Hemmsen, T. Kjeldsen, L. Larsen, C. Kjaer, B. G. Tomov, J. Mosegaard, and J. A. Jensen, “Implementation of synthetic aperture imaging on a hand- held device,” in Proceedings of 2014 ieee international ultrasonics symposium , 2014, pp. 2177-2180. 19/03/15 Page 5

Medical Ultrasound Click to edit Master title style http://www.bkmed.com/products_en.htm 19/03/15 6

Mobile transducer Click to edit Master title style

Leveraging disruptive technology Click to edit Master title style 19/03/15 8

Ultrasound Click to edit Master title style • From acoustic (pressure) waves to images Velocity (m/ Medium sec) Fat 1450 Water 1480 Soft tissue 1540 Kidney 1560 Blood 1570 Muscle 1580 Bone 4080 19/03/15 9 http://www.sensorwiki.org/doku.php/sensors/ultrasound

Beamforming to reconstruct images Click to edit Master title style 19/03/15 10

Shooting for a number of scanlines Click to edit Master title style 19/03/15 Page 11

Dynamic Receive beamforming Click to edit Master title style Multiple transducer elements Scanline Dynamic focus point 19/03/15 19/03/15 Page 12 Page 12

Beamforming Click to edit Master title style • Traditional beamforming requires a high data bandwidth. • A typical system could have 128 channels and use a 12- bit 40 MHz sampling system. • This generates 128 × 40 × 10 6 Hz × 2B = 9.54 GB/s 19/03/15 Page 13

SASB – dual beamforming Click to edit Master title style • Simple first stage – Single focal point for both transmit and receive • Advanced second stage – combining information from multiple first stage focused scan lines • Reduction in data-transfer requirement – Reduced by a factor of receive elements (192) M. C. Hemmsen, J. M. Hansen, and J. A. Jensen, “Synthetic Aperture Sequential Beamformation applied to medical imaging using a multi element convex array transducer,” in EUSAR , Apr. 2012, pp. 34–37. 19/03/15 Page 14

Algorithmic engineering Click to edit Master title style 19/03/15 15

1. stage: Fixed focus transmit and receive Click to edit Master title style Receiver r receives at time a+2*b +c a c First stage line Fixed focus point b 19/03/15 Page 16

2. stage: reconstruct focus Click to edit Master title style 1. Find index scanline entries 2. Add contribution 19/03/15 Page 17

The math behind it Click to edit Master title style 19/03/15 Page 18

Pseudo code Click to edit Master title style for all image samples p with polar coordinates (i,j) for all first stage scanlines k a = calculateWeight (i,k) d = calculateDelay (i,k) s = getScanline (d,k) l(i,j) += a*s 19/03/15 Page 19

Three implementations Click to edit Master title style AVX Multithreaded 19/03/15 Page 20

Benchmark, simple implementation Click to edit Master title style Simple ¡ 1,000,000 ¡ 600,000 ¡ 100,000 ¡ 10,000 ¡ ms ¡per ¡frame ¡ 2950 ¡ 1,000 ¡ 699 ¡ Simple ¡ 100 ¡ 10 ¡ 3.05 ¡ 1 ¡ 0 ¡ Matlab ¡ C ¡(1 ¡thread) ¡ C ¡(8 ¡threads) ¡ OpenCL ¡ Intel Core i7 2600 Nvidia GTX 680 GPU 19/03/15 Page 21

Algorithmic optimization Click to edit Master title style • Sampling the beamforming directly in the scan line sample locations 19/03/15 Page 22

Weights and delays precalculated Click to edit Master title style r k Delays r k Apodization 19/03/15 Page 23

Pseudo code Click to edit Master title style l = 0 for all image samples p with polar coordinates (i,j) for all first stage scanlines k - up to N(r i ) a = getWeight (i,k) if a=0 then break d = getDelay (i,k) s = getScanline (d,j+k) l(i,j) += a*s if (k>0) then s = getScanline (d,j-k) l(i,j) += a*s 19/03/15 Page 24

Benchmark, Optimization Click to edit Master title style 1,000,000 ¡ 600,000 ¡ 100,000 ¡ 10,000 ¡ ms ¡per ¡frame ¡ 2950 ¡ 1,000 ¡ Simple ¡ 699 ¡ 100 ¡ OpAmizaAon ¡ 10 ¡ 20.9 ¡ 5.4 ¡ 3.05 ¡ 1 ¡ 0.49 ¡ 0 ¡ Matlab ¡ C ¡(1 ¡thread) ¡ C ¡(8 ¡threads) ¡ OpenCL ¡ Intel Core i7 2600 Nvidia GTX 680 GPU 19/03/15 Page 25

Resulting image quality Click to edit Master title style Matlab SIMD/Multicore OpenGL OpenCL RMSE 0.0044 0.0042 0.0040 PSNR 47.23dB 47.79dB 48.01dB 19/03/15 Page 26

Going mobile (OpenGL ES 3.0) Click to edit Master title style 19/03/15 Page 27

Mobile hardware Click to edit Master title style LG G2 Samsung Samsung Nvidia Jetson HTC Nexus 9 Galaxy Tab Nexus 10 TK1 SoC Snadragon 800 Exynos 5 Exynos 5220 Tegra K1 Tegra K1 GPU Adreno 300 Mali T628 Mali T604 Kepler Kepler Screen 1920x1080 2560x1600 2560x1600 1920x1080 2048x1536 OS Android Android Android Linux4Tegra Android 19/03/15 Page 28

Mobile WIFI capabilities Click to edit Master title style • Need 25.3 MB/s à IEEE 802.11ac WIFI ¡throughput ¡ 50 ¡ 45 ¡ 40 ¡ 43.4 ¡ 35 ¡ 35 ¡ 30 ¡ MB/s ¡ 28.8 ¡ 25 ¡ 20 ¡ 15 ¡ 18.5 ¡ 10 ¡ 11.2 ¡ 5 ¡ 0 ¡ LG ¡G2 ¡ Galaxy ¡Tab ¡Pro ¡ Nexus ¡10 ¡ ¡ Jetson ¡TK1 ¡+ ¡Intel ¡ Nexus ¡9 ¡ 7260HMW ¡ 19/03/15 Page 29

Mobile performance Click to edit Master title style Timings ¡ 70 ¡ 60 ¡ 50 ¡ Scanconversion ¡ 40 ¡ BeamformaAon ¡ ms ¡ 30 ¡ IQ ¡demodulaAon ¡ Datatransfer ¡ 20 ¡ Total ¡Aming ¡ 10 ¡ 0 ¡ LG ¡G2 ¡(Adreno ¡330) ¡ Galaxy ¡Tab ¡Pro ¡(Mali ¡ Nexus ¡10 ¡(Mali ¡T604) ¡ Jetson ¡TK1 ¡(Tegra ¡K1) ¡ Nexus ¡9 ¡(Tegra ¡K1) ¡ T628) ¡ 19/03/15 Page 30

Going mobile, Nexus 9 Click to edit Master title style 19/03/15 Page 31

Digital or wireless? Click to edit Master title style 19/03/15 32

Click to edit Master title style Thank you for your attention Mail: jesper.mosegaard@alexandra.dk Twitter: @mosegaard LinkedIn: linkedin.com/in/mosegaard Please complete the Presenter Evaluation sent to you by email or through the GTC Mobile App. Your feedback is important!