Front-line research in Ultrasound CFU research Part 2: Fast Vector Flow Imaging, Pressure Gradients, Super Resolution and Portable Scanners

Transcription

1 Front-line research in Ultrasound CFU research Part 2: Fast Vector Flow Imaging, Pressure Gradients, Super Resolution and Portable Scanners Can measure 3-D flow fast with only 124 transducer elements? Jørgen Arendt Jensen Can we break the speed-accuracy trade-off? Can we make it in a portable scanner? Department of Electrical Engineering Can we measure pressure non-invasively? Can we see brain function and epileptic attack? Can we resolve structures below the resolution limit? Department of Electrical Engineering Systolic blood flow in ascending aorta of a patient with aortic valve stenosis (short axis view) Hansen et al. Ultrasonic Imaging, Vol. 35, No. 4, D Imaging and Matrix Probes O(,) x y z 32 x 32 = 124 element Matrix Probe from Vermon 1

2 Measurement equipment 3D Measurement situation Cross-sectional scan of vessel: Jensen et al. IEEE UFFC, No. 9, 213 2D transducer 32x32 elements 124 elements SARUS experimental scanner 124 channels 3D vector flow image 124 channels and 7 MHz/12 bits sampling for 14 Gbytes/s Arbitrary transmission of coded signals 32 high-end FPGAs for real-time processing More than 128 Gbytes RAM for several seconds of data Capable of 25.6 billion mults/s Can perform real-time SA imaging at 32 frames/s 2

3 3-D in-vivo flow in Carotid Artery Problem with Fully Populated Arrays 2D full array 124 channels, should be 496 Expensive! New row-column Array prototype 124 channels Less expensive Row-Column Addressed Array Reduction of the Number of Channels Fully Addressed Row-Column Addressed N N N For N = 62 Total elements = 3,844 N For N = 62 Total elements = 124 N 2 elements High channel count Reduction of N/2 = 31 times! N+N elements Low channel count 3

4 Advantages of Row-Column Arrays Matrix Array Focusing is dependent on N 2 Resolution: FWHM = λ D/W = λ D/(N pitch) Double resolution leads to 4 times number of elements Penetration depth depends on area Many elements needed Sparse arrays suffer in penetration RC Array Focusing dependent on N Resolution scales linearly Double resolution double number of elements Penetration depends on area Area = (N pitch) 2 Double number of elements gives four times area Row-Column Imaging 1 N array N+N row-column addressed array Horizontal Array Vertical Array Connection Element N 1 array SNR and Penetration Depth Video of RC Imaging Bouzari et al, SPIE MI, 216 4

5 Steered Emission Sequence Pulsatile Flow Steered Emission Sequence Pulsatile Flow Mean flow rate during one cycle 7 Mean vel. ± 1 std Flow rate [ml/s] 6 x z 5 Est. : 2.3 ±.1 ml/stroke True: 2.5 ml/stroke y z Holbek et al, IEEE UFFC, Vol. 63, No. 11, Time [s] Holbek et al, IEEE UFFC, Vol. 63, No. 11, 216 Volumetric 3-D Vector Flow Synthetic aperture and plane wave velocity imaging Department of Electrical Engineering 5

6 Limitations of current ultrasound systems Synthetic aperture imaging Only one transmit focus Frame rate is limited especially for blood flow and 3D imaging c f = 2DN N 154 = = 6 r d l Hz Velocity estimation is poor due to few data samples Velocity estimation is angle dependent - no transverse velocity can be estimated Performance Using 8.5 MHz Linear Array Transducer Synthetic Aperture Flow Imaging Dynamic Range: 5 db ~5% increase in penetration depth Lateral resolution performance better after aperture is fully opened at ~55mm Axial resolution improved due to better utilization of bandwidth 6

7 In-vivo SA Flow Image Parabolic Profiles from Flow Rig Parabolic velocity profile 24 emissions used 5 used in traditional systems -> 2 times faster Standard deviation relative to maximum: 3% More than 2 images/sec to 15 cm depth 9 images/sec at 4 cm Only 24 emissions for whole image The Royal Swedish Academy of Sciences Perspectives in Medical Diagnostics and Therapy Villagomez et al, IUS 215 SA Vector Flow Imaging in Carotid bifurcation Frame Rate: 2 Hz The jugular vein and carotid artery 27 7

8 Diagnostic value of pressure gradients [surgery.ucsf.edu] The jugular vein and carotid artery 3 Intravascular Pressure Drop along a Streamline Estimating pressure gradients from Navier-Stokes equation Pressure gradient Standard Deviation: 4% Neglect gravitation as patient lies down Temporal acceleration $ v ' ρ & + v v) = p + ρ g + µ 2 v % t ( m a = F Spatial acceleration Neglect viscosity due to large vessels The TO method can estimate both spatial and temporal acceleration 8

9 In-vivo scanning of carotid artery Non-invasive mapping of pressure in-vivo 33 Ultrasound in the Pocket Everywhere 34 Data Processing in an Ultrasound Scanner Plane wave and SA flow imaging demands a lot of calculation. 2 GB/sec Can we make an approach that can be put on a Tablet? 31 MB/sec Wireless transmission: - USB 2: 6 MB/sec n: 37 MB /sec 9

10 Synthetic Aperture Sequential Beamforming First stage fixed-focus beamformer Second stage dynamic beamformer SASB explained Virtual sources (VS) are created using a fixed focal depth in transmit and receive. Scan lines contain information from a set of spatial positions limited by the opening angle of the VS. A single image point is potentially represented in multiple scan lines. A second beamformer creates high resolution image points from multiple first stage lines. 1) J. Kortbek, J. A. Jensen, and K. L. Gammelmark, Synthetic aperture sequential beamforming, Ultrasonics SASB improves image quality compared to conventional imaging. SASB decreases data rate from probe to system from 5 Gbytes/s to 3 Mbytes/s SASB TO scan sequence Transverse lines beamforming D k l lateral signal x hr 1

11 Performance of SASB VFI Bias: -.67% - STD: 3.19% Lateral position [mm] Normalized amplitude 4 fixed focused emissions beamformed and combined to one HR line 1.5 Line is Hilbert transformed to give complex signal Signal Hilbert DTO estimation used for finding velocity vector Bias: -.43% - STD:.47% Axial velocity [m/s] Axial position [mm] 49 Lateral velocity [m/s] High-resolution double-oscillating point spread function Mean +/- 3std Reference Axial position [mm] Lateral position [mm] Spatial frequency [1/m] Velocity [m/s].8 A Time [sec].8 Velocity [m/s] Amplitude [db] -2 Align B Time [sec] 11

12 System Setup Asus rt-87u Evaluation of the Wireless Throughput Flow rig BK3 The required data throughput for real-time imaging is 13 MB/s. Yields 26 frames/s HTC Nexus 9 External cooler Nexus 9 tablet from 214, Nvidia Tegra K1 Very low velocity flow Properties of SASB VFI Data can be streamed wirelessly from the probe at 13 Mbytes/s Vasculatur Velocity map Beamforming in Tablet or cloud DTO vector flow More than 35 independent frames/s Up to 214 frames/s for measurements Data continuously available at all image points for all time for measurements 7% standard deviations in-vivo Mace et al.: Functional ultrasound imaging of the brain, Nature Methods 8, (211) 12

13 Brain Disorders and Nervous Systems Epileptic seizure in a rat brain CBV Changes (%) Courtesy of: E. Macé, G. Montaldo, I. Cohen, M. Baulac, M. Fink, M. Tanter, Nature Methods, July 211 Super Resolution Image Super resolution images can be obtained by compensating for tissue motion Image from: Ultrafast ultrasound localization microscopy for deep superresolution vascular imaging, Claudia Errico et al. Nature, November

14 Microbubble tracking for super resolution imaging Microbubble Characteristics Diffraction -> Point Spread Function Weighted centroid -> Exact Position Phantom tube, 2 µm Microbubbles injected Axial [mm] Microbubble infusion Lateral [mm] *FWHM = Full Width at Half Maximum In Vivo Experiment B-mode Sequence Direct microbubble infusion Left kidney Transducer Sequence parameter Value Pulse repetition frequency 28 khz Center frequency 12 MHz Framerate 5 Hz Axial resolution 25.5 µm Lateral resolution 43. µm Physical conditions Respiration frequency Rat pulse frequency Value 72 Hz 35 Hz 14

15 Motion Tracking Without motion compensation σgreen = 22 µm σblue = 34 µm With motion compensation σgreen = 8 µm σblue = 13.5 µm σ = 1.7 µm Inflow Outflow Inflow Outflow 15

16 What we can do with ultrasound Vector velocity can be measured fast and in all directions Velocity magnitude correctly identified No angle correction correct magnitude and direction for all times and places Disturbed and turbulent flow can be correctly visualized in 3-D New methods: SA and plane wave solves the speed-accuracy limit and can be in a portable scanner Center for Fast Ultrasound imaging Pressure gradients can be deduced from the vector flow Brain function can be studied Flow in the smallest vessels can be visualized DTU Elektro+Nanotech BK Medical Rigshospitalet Panum Many student projects on these topics Tour in the lab 16