- Email: [email protected]
A novel two-axis micromechanical scanning transducer using water-immersible electromagnetic actuators for handheld 3D ultrasound imaging
Sensors and Actuators A 236 (2015) 281–288
Contents lists available at ScienceDirect
Sensors and Actuators A: Physical journal homepage: www.elsevier.com/locate/sna
A novel two-axis micromechanical scanning transducer using water-immersible electromagnetic actuators for handheld 3D ultrasound imaging Chih-Hsien Huang, Jun Zou ∗ Department of Electrical and Computer Engineering, Texas A&M University, College Station, TX 77843, USA
a r t i c l e
i n f o
Article history: Received 23 June 2015 Received in revised form 30 August 2015 Accepted 18 September 2015 Available online 28 October 2015 Keywords: Scanning transducer Water-immersicible microactuator 3D ultrasound imaging
a b s t r a c t This paper reports the development of a new two-axis micromechanical scanning transducer for handheld 3D ultrasound imaging. It consists of a miniaturized single-element ultrasound transducer driven by a unique 2-axis liquid-immersible electromagnetic microactuator. With a mechanical scanning frequency of 19.532 Hz and an ultrasound pulse repetition rate of 5 kHz, the scanning transducer was scanned along 60 concentric paths with 256 detection points on each to simulate a physical 2D ultrasound transducer array of 60 × 256 elements. Using the scanning transducer, 3D pulse-echo ultrasound imaging of two silicon discs immersed in water as the imaging target was successfully conducted. The lateral resolution of the 3D ultrasound image was further improved with the synthetic aperture focusing technique (SAFT). The new two-axis micromechanical scanning transducer does not require complex and expensive multichannel data acquisition (DAQ) electronics. Therefore, it could provide a new approach to achieve compact and low-cost 3D ultrasound imaging systems, especially for handheld operations. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Ultrasound imaging (2D or 3D) has become a useful nondestructive diagnostic technique with a wide range of applications [1,2]. To conduct 3D ultrasound imaging, the time-variant ultrasound field at a 2D array of locations has to be properly recorded for image reconstruction [3–6]. Currently, there are three different methods to achieve 3D ultrasound imaging. First, a 2D ultrasound transducer array can be used to detect the incoming ultrasound signals [7–9]. In a 2D ultrasound transducer array, the transducer elements are usually arranged into an orthogonal matrix. Each transducer element or a sub-group of elements is interfaced with a data acquisition (DAQ) channel to send a short interrogating ultrasound pulse and also receive the time-variant backscattered signal (“echo”). The location of the transducer elements in the 2D array provides the lateral (x and y) information of the ultrasound scatterers in the imaging target, while the travel time of the echo signals is used to retrieve their depth (z) information. As a result, an ultrasound image with 3D scattering contrast can be reconstructed on a computer. The imaging resolution depends on the density and the working frequency of the transducer elements, while the field-of-
∗ Corresponding author. E-mail address: [email protected] (J. Zou). http://dx.doi.org/10.1016/j.sna.2015.09.025 0924-4247/© 2015 Elsevier B.V. All rights reserved.
view is determined by the overall size of the 2D transducer array. To obtain good imaging resolution (e.g., 1 mm) and field-of-view (e.g., a few cm), a large number of transducer elements (e.g., 1000s) and DAQ channels (e.g., 100s) are required. As a result, the entire imaging system could become complex, bulky, power-consuming, and expensive [10]. To address this issue, a 1D transducer array could be used to conduct “electronic” 2D B-Scan, while the scan in the azimuth dimension is conducted mechanically by using a one-axis motor stage or just by hand with the assistance of a position tracking device [11,12]. However, this method still requires an ultrasonic transducer array and multi-channel DAQ electronics. The need of a position tracking device complicates the imaging system design and operation [13,14]. Alternatively, the ultrasound signals can also be received by mechanically scanning a single-element transducer over the imaging target by using a two-axis motor stage. However, the use of 2-axis motor stages makes the entire imaging system complex and bulky. Second, the slow mechanical scanning frequency limits the date acquisition speed. As a result, this technique is mainly limited for lab use and is not suitable for handheld operations [15–17]. In this paper, we report a new 2-axis micromechanical scanning transducer technique to enable fast and versatile 3D ultrasound imaging. The 2-axis micromechanical scanning transducer consists of a miniaturized single-element transducer mounted a unique 2axis water-immersicible electromagnetic micro actuator. When AC
282
C.-H. Huang, J. Zou / Sensors and Actuators A 236 (2015) 281–288
is fixed onto a flexible hinge structure with a permanent magnet attached onto it. To provide the driving force for scanning the transducer in two axes, two pairs of inductor coils are mounted close to the permanent magnet. The magnetic polarity of the two inductor coils in each pair is made opposite. When an AC current is flowing through the two inductor coils, a push and a pull force will be generated on the permanent magnet to vibrate the flexible hinge together with the transducer at a certain scanning angle (). The magnetic force (F) generated between the permanent magnets and the inductor coils can be determined by F = V × Ms ×
∂H ∂z
––– (1),
where V is the total volume of the permanent magnet, Ms is its effective magnetization, and H is the magnetic field intensity generated by the inductor. Since the length of the hinge is much larger than that of the permanent magnet, the magnetic force can be considered as a point force and the resulting scanning angle () can be determined by =
Fig. 1. (a) Schematic of the 2-axis micromechanical scanning transducer design. (b) Picture of the constructed prototype.
driving currents are applied, the single-element transducer can be scanned along concentric paths with different radii at a frequency of 10–100 Hz. By synchronizing the scanning frequency with the ultrasound pulse-echo repetition rate, the ultrasound signals from a 2D array of locations can be received sequentially with only one T/R channel for 3D image reconstruction. Compared with the scanning with motor stages, the two-axis micromechanical scanning transducer can achieve higher scanning speed and therefore imaging speed with lower power consumption, which is due to its smaller scan mass. More importantly, it can be further miniaturized to be fitted into a small liquid-filled probe for handheld operations. Therefore, it could provide a new approach for compact, fast and low-cost 3D ultrasound imaging systems. 2. Design, construction and characterization 2.1. Design and construction To enable 3D ultrasound imaging, scanning the single-element transducer in two axes is needed. Since ultrasound waves with a frequency in the MHz range have high attenuation in air (1.64 dB/MHz-cm @ 20 ◦ C), a liquid coupling medium with low acoustic attenuation such as water (0.0022 dB/MHz-cm) is needed for their effective propagation. Therefore, a micro actuator that can work properly under water is desirable. Such that both the single-element scanning transducer and the micro actuator could be packaged into a liquid (water)-filled imaging probe, which is suitable for handheld operations along different orientations. Currently, the most commonly used micro actuators are piezoelectric actuators. However, they require high driving voltages (e.g., >75 V [18]), which could cause electrical breakdown or shorting in water. Their work distance is also limited (e.g., micrometers). In contrast, electromagnetic actuators do not need such high voltages to drive, which are able to work in a liquid environment. Besides, their work distance can reach millimeters or even centimeters. Therefore, electromagnetic actuation will be chosen as the driving mechanism for the scanning transducer. Fig. 1(a) shows the schematic design of the 2-axis micromechanical scanning transducer. A miniaturized single-element transducer
FL2 2EI
––– (2),
where L, E and I are the length, effective Young’s modulus and bending moment of inertia of the hinge, respectively. When driven with an AC current, the scanning motion can be described as a simple harmonic vibration. Its resonance frequency in air can be estimated by
fr
air
=
k m
1 2
––– (3),
where k is the bending force constant of the hinge and m is the overall effective mass of the scanning transducer assembly. When the scanning transducer is immersed in water, the resonant frequency (fr water ) can be estimated by fr
water
= fr
air
2
1 − 2␦
––– (4),
where ␦ is the effective damping ratio of the scanning transducer in water. Fig. 1(b) shows the constructed prototype of the 2-axis micromechanical scanning transducer. As the initial demonstration, a miniaturized water-immersion single-element transducer (XMS-310-B, Olympus) was used as the scanning transducer. It has a center frequency of 10 MHz, a 6-dB bandwidth of 80%, and a diameter of 2 mm. However, to meet the actual imaging requirements, other single-element transducers can be used as well. The RF coaxial cable of the single-element transducer was directly used as the bending hinge, which was clamped onto a height adjustable stage. A neodymium ring magnet (R84 × 0, K&J Magnetics) was used as the permanent magnet. It has a length of 10 mm, an outer diameter of 5 mm, and an inner diameter of 2 mm, respectively. Its nominal peak magnetic field intensity is ∼13200 Gauss. Eight RF coil inductors (70F331AF-RC, Bourns) were used as the driving coils. The inductance of each inductor is 330 mH. To provide the needed driving force, two inductors were connected in parallel. The use of two smaller coils instead of a larger one results in a more compact structure and more uniform field distribution. The inductors and their wire connections were coated with water-proof epoxy. All the components were assembled together with acrylic fixtures made by laser cutting. When the frequency of the AC driving current matches the resonance frequency, will reach its maximum, which results in the most efficient driving condition. Due to its centrosymmetric structure, the resonance frequencies in both axes are identical. Therefore, an ideal 2D scanning pattern will be a circular path (clockwise or counterclockwise), where both axes are driven at the
C.-H. Huang, J. Zou / Sensors and Actuators A 236 (2015) 281–288
283
Fig. 3. The scanning angle as a function of the AC driving voltage.
etition rate (fu ). Here, fu represents the total number of T/R cycles occurring in one second, while fr represents the number of circles along which the transducer is scanning in the same amount of time. Therefore, the ratio between fu and fr determines the number of T/R cycles (i.e., ultrasound detection points) occurring on each circular path. The time interval (t)between two adjacent detection points will be the repetition period of the ultrasound pulses. The relative location of each detection point along on the circular path can be determined sequentially along with the driving signal. To demonstrate the two-axis scanning, the optical trace of a LED-illuminated optical fiber tip driven by the two-axis microactuator is shown in Fig. 2(c). The pulser-receiver (5072PR, Olympus) used in this work can only provide a fixed pulse repetition rate of 5 kHz. To generate 256 detection points on one circular path, the resonance frequency will need to be adjusted to 5000/256 = 19.5315 Hz. For the data acquisition, after 256 ultrasound signals are recorded, the scanning of the transducer will be moved onto a different circular path by adjusting the amplitude of the driving signal. The data recording will start again after a delay period ().
2.2. Scanning characterization and adjustment
Fig. 2. (a) Waveforms of the two AC driving signals and ultrasound pulses. (b) The resulting distribution of ultrasoud detection points on a circular scanning path. (c) Picture of the optical trace of a LED-illuminated optical fiber tip driven by the twoaxis microactuator.
resonance frequency [18–20]. As shown in Fig. 2(a), to generate a circular scanning path, the two AC driving currents must follow Ix (t) = A cos(r t) = A cos(2fr t)
––– (5),
Iy (t) = A sin(ωr t) = A sin(2fr t)
––– (6),
where A is the amplitude of the driving currents and fr is the resonance frequency of the hinge and transducer assembly [21]. By adjusting the amplitude of the driving currents, the transducer can be scanned along one circular path with different radius (r), which is proportional to A (Fig. 2(b)). The location of the transducer at a certain time ti can be represented as (rcos (ωr ti ), rsin (ωr ti )). To dynamically configure the location of the ultrasound detection points on each circular path, the mechanical scanning of the transducer needs to be synchronized with the ultrasound pulse rep-
The entire scanning transducer setup was immersed in water to characterize the scanning angle when different driving currents were applied. Two function generators (33,220A and 33,220B, Agilent) were used to generate two AC signals at a frequency of 19.532 Hz, which have the same amplitude but with a phase difference of 90◦ . They were further amplified by two home-made current amplifiers before being sent to the inductor coils. Since the impedance of the inductors remains constant under the same frequency, the voltages applied on the inductors should be proportional to the currents flowing into them. Therefore, during the scanning characterization, the AC voltage drop on the inductor coils was monitored as a measure of the strength of the driving signals. Before the characterization of the scanning angle was conducted, the resonance frequency of the scanning transducer setup was finetuned to around 19.5 Hz. A driving voltage of 2.5 V (peak to peak) was applied to the inductor coils. The length of the supporting hinge was carefully adjusted, so that the scanning angle reached its maximum. Next, the driving voltage was swept from 0 to 5 V (peak to peak) with an increment of 0.1 V. The amplitude of the vibration of the transducer was recorded and the scanning angle was calculated accordingly. The characterization was repeated on both axes. As shown in Fig. 3, the scanning angle changes almost linearly with the driving voltage. During the imaging experiment, this linear rela-
284
C.-H. Huang, J. Zou / Sensors and Actuators A 236 (2015) 281–288
Fig. 5. Schematic of the ultrasound transmission from a source point to the transducer elements for SAFT.
driving signal, since all the 8192 sampling points have the same values, only the first sampling point was saved. After the data from all the 256 detection points on one circular path were recorded, the amplitude of the two driving currents was adjusted to move the transducer onto the next one. In the end, a total of 60 × 256 detection points for the ultrasound signals and the driving signals were acquired and saved on the computer for image reconstruction. 3.2. Imaging reconstruction
Fig. 4. (a) Schematic of the data acquisition system. CHA and CHB are the input channels and EXT is the external trigger channel of the DAQ card. (b) Representative driving and ultrasound signals recorded at one detection point.
tionship was used to determine the amplitude of the driving voltage to control the radius of a circular scanning path. 3. Data acquisition and image reconstruction
To achieve a good balance between the lateral resolution and computation time, 2D SAFT was applied separately along two orthogonal directions rather than applying 3D SAFT directly [22–24]. SAFT is a digital imaging processing technique widely used in the ultrasound imaging field. Its basic idea is to retrieve the total time-variant ultrasound signal emitting from one source point (i.e., one pixel in the image) based on their traveling time (delay) from the source point to the transducer elements. As shown in Fig. 5 Suppose the ultrasound wave from one source point (i.e., one image pixel) reaches the ith transducer element after a certain delay (ti ), the total ultrasound signal from the observation point (U (t)) can be calculated as
i c
3.1. Data acquisition
U(t) = ni=1 Ui (t + ti ) = ni=1 Ui t +
A schematic of the data acquisition system is shown in Fig. 4(a). The pulser-receiver (5072PR, Olympus) was set to provide a pulse repetition rate of 5 kHz, an amplification of 50 dB, and a damping ratio of 20 . A personal computer loaded with LabView (version 2013, National Instruments) was used to trigger the two function generators (33,220A and 33,220B, Agilent) to generate two 19.532 Hz (which is the closest value to 19.53125 Hz provided by the function generators) sinusoidal driving signals with 90◦ phase shift. After being amplified by the current amplifiers, the two driving currents were sent to the scanning actuator to create a circular scanning path of the transducer. The diameter of each circular path is determined by the amplitude of the driving currents or voltages (Fig. 3). To achieve 60 × 256 detection points, the transducer was scanned on 60 circular paths with equal intervals. The external trigger port of the DAQ card (AT9350-128 M, Alazar Technologies) was connected to the synchronization output port of pulser-receiver. Once being triggered by the pulser-receiver, it will start the data collection with 100 MHz sampling rate and 8192 sampling points. As a result, the time interval between two sampling points is 0.01 s, and the total acquisition time for each ultrasound signal is 81.92 s. Although the transducer is continuously moving, the location of the transducer can be considered static at each detection point due to the short data acquisition time. At each detection point (Fig 2(b)), both the amplified ultrasound echo signals (from the pulser-receiver) and the driving signals from the function generator were recorded (Fig. 4(b)). For each ultrasound signal, all the 8192 sampling points were saved. For each
where n is the total number of transducer elements in the array, Ui (t) is the ultrasound signal received by the ith transducer element, i is the observation distance between the source point and the ith transducer element, and c is the sound velocity. By repeating such calculation pixel by pixel, the ultrasound signal from each pixel on the 2D imaging (XY) plane can be “electronically focused”, and the contrast and lateral resolution of the reconstructed ultrasound image can be enhanced. As shown in Fig. 6(a) and (b), the cone-shaped 3D imaging space were decomposed into 60 concentric circular planes and 128 sector planes, which are perpendicular to each other. For each circular plane, a raw B-Scan image was reconstructed from the 256 A-lines (the ultrasound signals received at the 256 detection points on the circular scan path). Since each A-line has 8192 sampling points, the B-Scan image consists of a total number of 256 × 8192 pixels. For each pixel, its horizontal (lateral) location is determined by the location of the corresponding detection point, while its vertical (axial) location is determined by its travel time multiplied by the sound velocity. The grayscale or the color of the pixel represents the strength of the ultrasound signal (i.e., the recorded value at the sampling point). After this, a second B-Scan image was reconstructed based new ultrasound signal strengths determined by SAFT. Following the same procedure, for each sector plane, two B-scan images were reconstructed from the 120 A-lines (with 8192 sampling points in each) with and without applying SAFT, respectively. Next, the B-Scan images for the 60 circular planes and the 128 sector planes were normalized and combined together pixel by
––– (1),
C.-H. Huang, J. Zou / Sensors and Actuators A 236 (2015) 281–288
Fig. 6. 3D ultrasound Image reconstruction process. (a) Circular B-Scan image reconstruction. (b) Sector B-Scan image reconstruction. (c) C-Scan image slicing. (d) 3D image formation.
285
Fig. 8. Picture of the imaging target.
d = ı cos
––(9),
where d is the distance between the two c-planes bounding Ti andPj , respectively, r is the diameter of the circular path passing the detection point (Ti ), r is the diameter of the circular path passing the reconstruction pixel (Pj ), is the scanning angle of the transducer, and ϕ is the angle between the two sector planes bounding Ti and Pj , respectively. For applying SAFT on a sector plane (Fig. 7(b)), the observation distance can be calculated as =
l2 l2 − 2ll cos
(6)
Where l is the distance from the ultrasound detection point (Ti ) to the anchor point of the hinge, l is the distance from the reconstruction pixel (Pi ) to the anchor point of the hinge, and is the cross angle between Ti and Pj with respect to the anchor point. 4. Imaging experiments and results
Fig. 7. Illustration of the geometry for the calculation of the observation distance for SAFT. (a) On circular planes. (b) On sector planes.
pixel to reconstruct the 3D image reconstruction, which consists of a total number of 60 × 256 × 8192 voxels. The voxels were further regrouped onto 8192C-Scan images in Cartesian coordinates and saved into the png format (Fig. 6(c)). At last, the 8192C-Scan images were stacked together to create a 3D view in Volview (Kitware) (Fig. 6(d)). Applying 2D SAFT on the circular and sector planes is similar with that on Cartesian coordinates, except that the observation distance between a detection point and a reconstruction pixel needs to be calculated in a different way. As shown in Fig. 7(a), for applying SAFT on a circular plane, the observation distance () from an ultrasound detection point (Ti ) to a reconstruction pixel ( Pj) can be calculated as = c=
c 2 + d2
r 2 + r 2 − 2rr cos ϕ
––– (7), –– (8),
Fig. 8 shows the imaging target for the ultrasound imaging experiments with the scanning transducer. It consists of two silicon discs with a diameter of 6 mm, which are supported with four optical fibers with a diameter of 100 m. To fix the optical fibers in place, four acrylic plates with an array of positioning holes were made by laser cutting and assembled together with epoxy. The distances between the two silicon discs are 15 mm, 10 mm, and 20 mm, respectively. Due to their small diameter, the back-scattered ultrasound signals from the optical fibers will be weak and therefore would not interfere with those from the two silicon discs. During the ultrasound imaging experiment, the entire assembly was placed underneath the scanning transducer setup and completely immersed in water. 4.1. Alignment of the detection points To implement the 2D SAFT on the circular and sector planes, the ultrasound detection points need to be positioned on the 2D grid formed by the circular and sector planes. However, due to the random time delay ( ) between the trigger signals from the LabView controlled computer and the driving signals from the function generators, the ultrasound detection points on the circular scanning paths could be misaligned (Fig. 9(a)). In addition, when the scanning of the transducer is shift from one circular path to another, some ultrasound detection points could appear on the transition section between two circular paths. To address this issue, the data recording of the DAQ card was delayed by a time () (longer than ) to exclude those off-path detection points. The total recording period
286
C.-H. Huang, J. Zou / Sensors and Actuators A 236 (2015) 281–288
the circular scanning paths can be re-aligned based on the phase of the corresponding driving signals (Fig. 9(b)). 4.2. 2D image reconstruction Fig. 10(a) shows the raw B-Scan image (before applying SAFT) of a representative circular plane reconstructed from the 256 Aline data. Fig. 10(b) shows the B-Scan image of the same circular plane after applying SAFT. The reverberation signals from the ultrasound echo traveled back and forth between sample and transducer were canceled by using the reverberation nulling and echo prediction method [25–27]. The signals reflected from the aluminum plate at the bottom of acrylic holder and from the tip of transducer itself were also removed. A similar image processing procedure was adopted for the section plane. Fig. 10(c) and (d) show the B-Scan images of one representative sector plane before and after applying SAFT. Both images consist of 120 A-line data. 4.3. 3D image reconstruction
Fig. 9. (a) Random time delay in the AC driving signals and the resulting misalignment of the ultrasound detection points. (b) Re-alignment of the ultrasound detection points based on the AC driving signals.
was still equal to the period of the driving signal. The relative location of an ultrasound detection point on a circular scanning path is determined by the time and therefore the phase of the AC driving signals (see Fig. 2). Therefore, the ultrasound detections points on
Fig. 11(a) shows the reconstructed 3D image based on the raw B-Scan images of both the circular and sector planes. The C-Scan images with gradually increased diameters can be clearly seen. The margin of the graphs excluding the Cartesian coordinates is presented in white color. In order to provide a better view of the two imaging targets, most of the margin was removed by using the shrink function in Volview (Fig. 11(b)). However, in order to keep the image of the two targets intact, a small portion of the white margin at the upper-left corner has to be kept and therefore is also shown in the plot. Fig. 11(c) shows the reconstructed 3D image based on the B-Scan images of both the circular and sector planes after applying SAFT. Compared with the image without
Fig. 10. B-Scan image of a representative (flattened) circular plane (a) before and (b) after applying SAFT. B-Scan image of a representative sector plane (c) before and (d) after applying SAFT.
C.-H. Huang, J. Zou / Sensors and Actuators A 236 (2015) 281–288
287
ing target was successfully conducted. The lateral resolution of the 3D ultrasound image was improved with modified SAFT algorisms based on the unique format of the acquired ultrasound data. The scanning transducer technique requires only one DAQ channel. Therefore it does not involve the complex and expensive multi-channel DAQ electronics in the ultrasound-array based imaging systems. Ultrasound imaging at different frequency ranges can be obtained by using a single-element transducer with different frequency response. The liquid-immersible electromagnetic microactuator provided 2-axis mechanical scanning at 10s–100s Hz. This made the acquisition of all the ultrasound data for 3D image reconstruction to be performed in a timely manner. This work laid a foundation for the development of compact and low-cost hand-held scanning transducer probes for 3D ultrasound imaging. In the future, new functionalities (e.g., optical delivery) will be investigated to expand the imaging capability (e.g., photoacoustic tomography) of the scanning transducer technique.
References
Fig. 11. Reconstructed 3D ultrasound images in Volview. (a) Raw image. (b) Image with white margin removed. (c) Image with white margin removed and after applying SAFT. The red, green and blue lines indicate the x y and z axes.
SAFT, both the contrast and spatial resolution of the imaging targets were significantly improved. The field of view of the scanning transducer is a cone shape (3-cm diameter on the top, 6-cm diameter circle at the bottom, and 6-cm depth). The lateral resolution is ∼1 mm, which is limited by the diameter (2 mm) of the transducer. The mean diameters of the upper and lower silicon discs were estimated to be 6.28 mm and 5.83 mm, respectively, which close to the actual dimension (6 mm). The distances between the two silicon pieces along x, y, and z axes were estimated to be 14.2 mm, 9.88 mm, and 20.11 mm, respectively, which match well with their actual values. 5. Conclusion In this work, a new two-axis micromechanical scanning transducer technique was developed and 3D pulse-echo ultrasound imaging of two silicon discs immersed in water as the imag-
[1] P.l. Greguss, Ultrasonic Imaging: Seeing by Sound: the Principles and Widespread Applications of Image Formation by Sonic, Ultrasonic, and Other Mechanical Waves, 1st ed., Focal Press, London, New York, 1980. [2] T.L. Szabo, Diagnostic Ultrasound Imaging: Inside Out, Elsevier Academic Press, Amsterdam, Boston, 2004. [3] A. Fenster, D.B. Downey, Three-dimensional ultrasound imaging, Annu. Rev. Biomed. Eng. 2 (2000) 457–475. [4] A. Fenster, D.B. Downey, 3-D ultrasound imaging: a review, IEEE Eng. Med. Biol. Mag. 15 (November–December) (1996) 41–51. [5] J.P.S. Garcia, P.T. Garcia, R.B. Rosen, P.T. Finger, A 3-dimensional ultrasound C-scan imaging technique for optic nerve measurements, Ophthalmology 111 (June) (2004) 1238–1243. [6] Q.H. Huang, Y.P. Zheng, M.H. Lu, Z.R. Chi, Development of a portable 3D ultrasound imaging system for musculoskeletal tissues, Ultrasonics 43 (January) (2005) 153–163. [7] E.D. Light, R.E. Davidsen, J.O. Fiering, T.A. Hruschka, S.W. Smith, Progress in two-dimensional arrays for real-time volumetric imaging, Ultrason. Imaging 20 (January) (1998) 1–15. [8] G. Rizzatto, Evolution of ultrasound transducers: 1.5 and 2D arrays, Eur. Radiol. 9 (Suppl. 3) (1999) S304–S306. [9] I.O. Wygant, X. Zhuang, D.T. Yeh, O. Oralkan, A. Sanli Ergun, M. Karaman, et al., Integration of 2D CMUT arrays with front-end electronics for volumetric ultrasound imaging, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 55 (February) (2008) 327–342. [10] F. Forsberg, Ultrasonic biomedical technology; marketing versus clinical reality, Ultrasonics 42 (April) (2004) 17–27. [11] R.N. Rohling, A.H. Gee, L. Berman, Automatic registration of 3-D ultrasound images, Ultrasound Med. Biol. 24 (July) (1998) 841–854. [12] Q.H. Huang, Y.P. Zheng, R. Li, M.H. Lu, 3-D measurement of body tissues based on ultrasound images with 3-D spatial information, Ultrasound Med. Biol. 31 (December) (2005) 1607–1615. [13] V. Grau, H. Becher, J.A. Noble, Registration of multiview real-time 3-D echocardiographic sequences, IEEE Trans. Med. Imaging 26 (September) (2007) 1154–1165. [14] X. Ye, J.A. Noble, D. Atkinson, 3-D freehand echocardiography for automatic left ventricle reconstruction and analysis based on multiple acoustic windows, IEEE Trans. Med. Imaging 21 (September) (2002) 1051–1058. [15] A. Briggs, O.V. Kolosov, Acoustic Microscopy, 2nd ed., Oxford University Press, Oxford, 2010. [16] R.G. Maev, Advances in acoustic microscopy and high resolution imaging, Wiley-VCH, John Wiley distributor, Weinheim, Chichester, 2013. [17] A.S. Savoia, G. Caliano, M. Pappalardo, A CMUT probe for medical ultrasonography: from microfabrication to system integration, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 59 (June) (2012) 1127–1138. [18] M.T. Myaing, D.J. MacDonald, X.D. Li, Fiber-optic scanning two-photon fluorescence endoscope, Opt. Lett. 31 (April) (2006) 1076–1078. [19] C.M. Lee, C.J. Engelbrecht, T.D. Soper, F. Helmchen, E.J. Seibel, Scanning fiber endoscopy with highly flexible, 1mm catheterscopes for wide-field, full-color imaging, J. Biophotonics 3 (June) (2010) 385–407. [20] G.J. Tearney, S.A. Boppart, B.E. Bouma, M.E. Brezinski, N.J. Weissman, J.F. Southern, et al., Scanning single-mode fiber optic catheter-endoscope for optical coherence tomography, Opt. Lett. 21 (April) (1996) 543–545. [21] E.J. Seibel, Q.Y.J. Smithwick, C.M. Brown, P.G. Reinhall, Single fiber flexible endoscope: general design for small size, high resolution, and wide field of view, Biomonitoring Endosc. Technol. 1 (2001) 29–39. [22] Z.L. Deng, X.Q. Yang, H. Gong, Q.M. Luo, Two-dimensional synthetic-aperture focusing technique in photoacoustic microscopy, J. Appl. Phys. 109 (May) (2011).
288
C.-H. Huang, J. Zou / Sensors and Actuators A 236 (2015) 281–288
[23] C.H. Frazier, W.D. O’Brien, Synthetic aperture techniques with a virtual source element, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 45 (January) (1998) 196–207. [24] N.V. Ruiter, G.F. Schwarzenberg, M. Zapf, H. Gemmeke, Conclusions from an Experimental 3D Ultrasound Computer Tomograph, 2008 IEEE Nuclear Science Symposium and Medical Imaging Conference (2008 Nss/Mic), vols. 1–9, pp. 3776–3783, (2009). [25] T. Folegot, C. Prada, M. Fink, Resolution enhancement and separation of reverberation from target echo with the time reversal operator decomposition, J. Acoust. Soc. Am. 113 (June) (2003) 3155–3160. [26] S.G. Parulekar, Ultrasonic detection of calcification in gallstones: the reverberation shadow, J. Ultrasound Med. 3 (March) (1984) 123–129. [27] H.C. Song, W.S. Hodgkiss, W.A. Kuperman, K.G. Sabra, T. Akal, M. Stevenson, Passive reverberation nulling for target enhancement, J. Acoust. Soc. Am. 122 (December) (2007) 3296–3303.
Biographies
Chih-Hsien HuangChih-Hsien Huang received his B.S. and M.S. degrees in Electrical Engineering from the National Cheng Kung University, Tainan, Taiwan, in 2003 and 2005, respectively. From 2005 to 2010, he was with Industrial Technology Research Institute—A Research and Development Foundation, Tainan, Taiwan, where he developed various feedback control methodologies for LED backlight system including current mixing architecture, flick free sequential detection method, and wavelength stabling algorithm. He is currently a Ph.D. student in the Department of Electrical and Computer Engineering at Texas A&M University. His research interest is focused on MEMS based scanning devices for acoustic tomography applications.
Jun Zou received his Ph.D. degree in electrical engineering from the University of Illinois at Urbana-Champaign in 2002. In 2004, he joined in the department of electrical and computer engineering at Texas A&M University, where he is currently an associate professor. His current research interests lie in the development of micro and nano optoelectro-mechanical devices and systems for biomedical imaging and sensing applications.
Contents lists available at ScienceDirect
Sensors and Actuators A: Physical journal homepage: www.elsevier.com/locate/sna
A novel two-axis micromechanical scanning transducer using water-immersible electromagnetic actuators for handheld 3D ultrasound imaging Chih-Hsien Huang, Jun Zou ∗ Department of Electrical and Computer Engineering, Texas A&M University, College Station, TX 77843, USA
a r t i c l e
i n f o
Article history: Received 23 June 2015 Received in revised form 30 August 2015 Accepted 18 September 2015 Available online 28 October 2015 Keywords: Scanning transducer Water-immersicible microactuator 3D ultrasound imaging
a b s t r a c t This paper reports the development of a new two-axis micromechanical scanning transducer for handheld 3D ultrasound imaging. It consists of a miniaturized single-element ultrasound transducer driven by a unique 2-axis liquid-immersible electromagnetic microactuator. With a mechanical scanning frequency of 19.532 Hz and an ultrasound pulse repetition rate of 5 kHz, the scanning transducer was scanned along 60 concentric paths with 256 detection points on each to simulate a physical 2D ultrasound transducer array of 60 × 256 elements. Using the scanning transducer, 3D pulse-echo ultrasound imaging of two silicon discs immersed in water as the imaging target was successfully conducted. The lateral resolution of the 3D ultrasound image was further improved with the synthetic aperture focusing technique (SAFT). The new two-axis micromechanical scanning transducer does not require complex and expensive multichannel data acquisition (DAQ) electronics. Therefore, it could provide a new approach to achieve compact and low-cost 3D ultrasound imaging systems, especially for handheld operations. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Ultrasound imaging (2D or 3D) has become a useful nondestructive diagnostic technique with a wide range of applications [1,2]. To conduct 3D ultrasound imaging, the time-variant ultrasound field at a 2D array of locations has to be properly recorded for image reconstruction [3–6]. Currently, there are three different methods to achieve 3D ultrasound imaging. First, a 2D ultrasound transducer array can be used to detect the incoming ultrasound signals [7–9]. In a 2D ultrasound transducer array, the transducer elements are usually arranged into an orthogonal matrix. Each transducer element or a sub-group of elements is interfaced with a data acquisition (DAQ) channel to send a short interrogating ultrasound pulse and also receive the time-variant backscattered signal (“echo”). The location of the transducer elements in the 2D array provides the lateral (x and y) information of the ultrasound scatterers in the imaging target, while the travel time of the echo signals is used to retrieve their depth (z) information. As a result, an ultrasound image with 3D scattering contrast can be reconstructed on a computer. The imaging resolution depends on the density and the working frequency of the transducer elements, while the field-of-
∗ Corresponding author. E-mail address: [email protected] (J. Zou). http://dx.doi.org/10.1016/j.sna.2015.09.025 0924-4247/© 2015 Elsevier B.V. All rights reserved.
view is determined by the overall size of the 2D transducer array. To obtain good imaging resolution (e.g., 1 mm) and field-of-view (e.g., a few cm), a large number of transducer elements (e.g., 1000s) and DAQ channels (e.g., 100s) are required. As a result, the entire imaging system could become complex, bulky, power-consuming, and expensive [10]. To address this issue, a 1D transducer array could be used to conduct “electronic” 2D B-Scan, while the scan in the azimuth dimension is conducted mechanically by using a one-axis motor stage or just by hand with the assistance of a position tracking device [11,12]. However, this method still requires an ultrasonic transducer array and multi-channel DAQ electronics. The need of a position tracking device complicates the imaging system design and operation [13,14]. Alternatively, the ultrasound signals can also be received by mechanically scanning a single-element transducer over the imaging target by using a two-axis motor stage. However, the use of 2-axis motor stages makes the entire imaging system complex and bulky. Second, the slow mechanical scanning frequency limits the date acquisition speed. As a result, this technique is mainly limited for lab use and is not suitable for handheld operations [15–17]. In this paper, we report a new 2-axis micromechanical scanning transducer technique to enable fast and versatile 3D ultrasound imaging. The 2-axis micromechanical scanning transducer consists of a miniaturized single-element transducer mounted a unique 2axis water-immersicible electromagnetic micro actuator. When AC
282
C.-H. Huang, J. Zou / Sensors and Actuators A 236 (2015) 281–288
is fixed onto a flexible hinge structure with a permanent magnet attached onto it. To provide the driving force for scanning the transducer in two axes, two pairs of inductor coils are mounted close to the permanent magnet. The magnetic polarity of the two inductor coils in each pair is made opposite. When an AC current is flowing through the two inductor coils, a push and a pull force will be generated on the permanent magnet to vibrate the flexible hinge together with the transducer at a certain scanning angle (). The magnetic force (F) generated between the permanent magnets and the inductor coils can be determined by F = V × Ms ×
∂H ∂z
––– (1),
where V is the total volume of the permanent magnet, Ms is its effective magnetization, and H is the magnetic field intensity generated by the inductor. Since the length of the hinge is much larger than that of the permanent magnet, the magnetic force can be considered as a point force and the resulting scanning angle () can be determined by =
Fig. 1. (a) Schematic of the 2-axis micromechanical scanning transducer design. (b) Picture of the constructed prototype.
driving currents are applied, the single-element transducer can be scanned along concentric paths with different radii at a frequency of 10–100 Hz. By synchronizing the scanning frequency with the ultrasound pulse-echo repetition rate, the ultrasound signals from a 2D array of locations can be received sequentially with only one T/R channel for 3D image reconstruction. Compared with the scanning with motor stages, the two-axis micromechanical scanning transducer can achieve higher scanning speed and therefore imaging speed with lower power consumption, which is due to its smaller scan mass. More importantly, it can be further miniaturized to be fitted into a small liquid-filled probe for handheld operations. Therefore, it could provide a new approach for compact, fast and low-cost 3D ultrasound imaging systems. 2. Design, construction and characterization 2.1. Design and construction To enable 3D ultrasound imaging, scanning the single-element transducer in two axes is needed. Since ultrasound waves with a frequency in the MHz range have high attenuation in air (1.64 dB/MHz-cm @ 20 ◦ C), a liquid coupling medium with low acoustic attenuation such as water (0.0022 dB/MHz-cm) is needed for their effective propagation. Therefore, a micro actuator that can work properly under water is desirable. Such that both the single-element scanning transducer and the micro actuator could be packaged into a liquid (water)-filled imaging probe, which is suitable for handheld operations along different orientations. Currently, the most commonly used micro actuators are piezoelectric actuators. However, they require high driving voltages (e.g., >75 V [18]), which could cause electrical breakdown or shorting in water. Their work distance is also limited (e.g., micrometers). In contrast, electromagnetic actuators do not need such high voltages to drive, which are able to work in a liquid environment. Besides, their work distance can reach millimeters or even centimeters. Therefore, electromagnetic actuation will be chosen as the driving mechanism for the scanning transducer. Fig. 1(a) shows the schematic design of the 2-axis micromechanical scanning transducer. A miniaturized single-element transducer
FL2 2EI
––– (2),
where L, E and I are the length, effective Young’s modulus and bending moment of inertia of the hinge, respectively. When driven with an AC current, the scanning motion can be described as a simple harmonic vibration. Its resonance frequency in air can be estimated by
fr
air
=
k m
1 2
––– (3),
where k is the bending force constant of the hinge and m is the overall effective mass of the scanning transducer assembly. When the scanning transducer is immersed in water, the resonant frequency (fr water ) can be estimated by fr
water
= fr
air
2
1 − 2␦
––– (4),
where ␦ is the effective damping ratio of the scanning transducer in water. Fig. 1(b) shows the constructed prototype of the 2-axis micromechanical scanning transducer. As the initial demonstration, a miniaturized water-immersion single-element transducer (XMS-310-B, Olympus) was used as the scanning transducer. It has a center frequency of 10 MHz, a 6-dB bandwidth of 80%, and a diameter of 2 mm. However, to meet the actual imaging requirements, other single-element transducers can be used as well. The RF coaxial cable of the single-element transducer was directly used as the bending hinge, which was clamped onto a height adjustable stage. A neodymium ring magnet (R84 × 0, K&J Magnetics) was used as the permanent magnet. It has a length of 10 mm, an outer diameter of 5 mm, and an inner diameter of 2 mm, respectively. Its nominal peak magnetic field intensity is ∼13200 Gauss. Eight RF coil inductors (70F331AF-RC, Bourns) were used as the driving coils. The inductance of each inductor is 330 mH. To provide the needed driving force, two inductors were connected in parallel. The use of two smaller coils instead of a larger one results in a more compact structure and more uniform field distribution. The inductors and their wire connections were coated with water-proof epoxy. All the components were assembled together with acrylic fixtures made by laser cutting. When the frequency of the AC driving current matches the resonance frequency, will reach its maximum, which results in the most efficient driving condition. Due to its centrosymmetric structure, the resonance frequencies in both axes are identical. Therefore, an ideal 2D scanning pattern will be a circular path (clockwise or counterclockwise), where both axes are driven at the
C.-H. Huang, J. Zou / Sensors and Actuators A 236 (2015) 281–288
283
Fig. 3. The scanning angle as a function of the AC driving voltage.
etition rate (fu ). Here, fu represents the total number of T/R cycles occurring in one second, while fr represents the number of circles along which the transducer is scanning in the same amount of time. Therefore, the ratio between fu and fr determines the number of T/R cycles (i.e., ultrasound detection points) occurring on each circular path. The time interval (t)between two adjacent detection points will be the repetition period of the ultrasound pulses. The relative location of each detection point along on the circular path can be determined sequentially along with the driving signal. To demonstrate the two-axis scanning, the optical trace of a LED-illuminated optical fiber tip driven by the two-axis microactuator is shown in Fig. 2(c). The pulser-receiver (5072PR, Olympus) used in this work can only provide a fixed pulse repetition rate of 5 kHz. To generate 256 detection points on one circular path, the resonance frequency will need to be adjusted to 5000/256 = 19.5315 Hz. For the data acquisition, after 256 ultrasound signals are recorded, the scanning of the transducer will be moved onto a different circular path by adjusting the amplitude of the driving signal. The data recording will start again after a delay period ().
2.2. Scanning characterization and adjustment
Fig. 2. (a) Waveforms of the two AC driving signals and ultrasound pulses. (b) The resulting distribution of ultrasoud detection points on a circular scanning path. (c) Picture of the optical trace of a LED-illuminated optical fiber tip driven by the twoaxis microactuator.
resonance frequency [18–20]. As shown in Fig. 2(a), to generate a circular scanning path, the two AC driving currents must follow Ix (t) = A cos(r t) = A cos(2fr t)
––– (5),
Iy (t) = A sin(ωr t) = A sin(2fr t)
––– (6),
where A is the amplitude of the driving currents and fr is the resonance frequency of the hinge and transducer assembly [21]. By adjusting the amplitude of the driving currents, the transducer can be scanned along one circular path with different radius (r), which is proportional to A (Fig. 2(b)). The location of the transducer at a certain time ti can be represented as (rcos (ωr ti ), rsin (ωr ti )). To dynamically configure the location of the ultrasound detection points on each circular path, the mechanical scanning of the transducer needs to be synchronized with the ultrasound pulse rep-
The entire scanning transducer setup was immersed in water to characterize the scanning angle when different driving currents were applied. Two function generators (33,220A and 33,220B, Agilent) were used to generate two AC signals at a frequency of 19.532 Hz, which have the same amplitude but with a phase difference of 90◦ . They were further amplified by two home-made current amplifiers before being sent to the inductor coils. Since the impedance of the inductors remains constant under the same frequency, the voltages applied on the inductors should be proportional to the currents flowing into them. Therefore, during the scanning characterization, the AC voltage drop on the inductor coils was monitored as a measure of the strength of the driving signals. Before the characterization of the scanning angle was conducted, the resonance frequency of the scanning transducer setup was finetuned to around 19.5 Hz. A driving voltage of 2.5 V (peak to peak) was applied to the inductor coils. The length of the supporting hinge was carefully adjusted, so that the scanning angle reached its maximum. Next, the driving voltage was swept from 0 to 5 V (peak to peak) with an increment of 0.1 V. The amplitude of the vibration of the transducer was recorded and the scanning angle was calculated accordingly. The characterization was repeated on both axes. As shown in Fig. 3, the scanning angle changes almost linearly with the driving voltage. During the imaging experiment, this linear rela-
284
C.-H. Huang, J. Zou / Sensors and Actuators A 236 (2015) 281–288
Fig. 5. Schematic of the ultrasound transmission from a source point to the transducer elements for SAFT.
driving signal, since all the 8192 sampling points have the same values, only the first sampling point was saved. After the data from all the 256 detection points on one circular path were recorded, the amplitude of the two driving currents was adjusted to move the transducer onto the next one. In the end, a total of 60 × 256 detection points for the ultrasound signals and the driving signals were acquired and saved on the computer for image reconstruction. 3.2. Imaging reconstruction
Fig. 4. (a) Schematic of the data acquisition system. CHA and CHB are the input channels and EXT is the external trigger channel of the DAQ card. (b) Representative driving and ultrasound signals recorded at one detection point.
tionship was used to determine the amplitude of the driving voltage to control the radius of a circular scanning path. 3. Data acquisition and image reconstruction
To achieve a good balance between the lateral resolution and computation time, 2D SAFT was applied separately along two orthogonal directions rather than applying 3D SAFT directly [22–24]. SAFT is a digital imaging processing technique widely used in the ultrasound imaging field. Its basic idea is to retrieve the total time-variant ultrasound signal emitting from one source point (i.e., one pixel in the image) based on their traveling time (delay) from the source point to the transducer elements. As shown in Fig. 5 Suppose the ultrasound wave from one source point (i.e., one image pixel) reaches the ith transducer element after a certain delay (ti ), the total ultrasound signal from the observation point (U (t)) can be calculated as
i c
3.1. Data acquisition
U(t) = ni=1 Ui (t + ti ) = ni=1 Ui t +
A schematic of the data acquisition system is shown in Fig. 4(a). The pulser-receiver (5072PR, Olympus) was set to provide a pulse repetition rate of 5 kHz, an amplification of 50 dB, and a damping ratio of 20 . A personal computer loaded with LabView (version 2013, National Instruments) was used to trigger the two function generators (33,220A and 33,220B, Agilent) to generate two 19.532 Hz (which is the closest value to 19.53125 Hz provided by the function generators) sinusoidal driving signals with 90◦ phase shift. After being amplified by the current amplifiers, the two driving currents were sent to the scanning actuator to create a circular scanning path of the transducer. The diameter of each circular path is determined by the amplitude of the driving currents or voltages (Fig. 3). To achieve 60 × 256 detection points, the transducer was scanned on 60 circular paths with equal intervals. The external trigger port of the DAQ card (AT9350-128 M, Alazar Technologies) was connected to the synchronization output port of pulser-receiver. Once being triggered by the pulser-receiver, it will start the data collection with 100 MHz sampling rate and 8192 sampling points. As a result, the time interval between two sampling points is 0.01 s, and the total acquisition time for each ultrasound signal is 81.92 s. Although the transducer is continuously moving, the location of the transducer can be considered static at each detection point due to the short data acquisition time. At each detection point (Fig 2(b)), both the amplified ultrasound echo signals (from the pulser-receiver) and the driving signals from the function generator were recorded (Fig. 4(b)). For each ultrasound signal, all the 8192 sampling points were saved. For each
where n is the total number of transducer elements in the array, Ui (t) is the ultrasound signal received by the ith transducer element, i is the observation distance between the source point and the ith transducer element, and c is the sound velocity. By repeating such calculation pixel by pixel, the ultrasound signal from each pixel on the 2D imaging (XY) plane can be “electronically focused”, and the contrast and lateral resolution of the reconstructed ultrasound image can be enhanced. As shown in Fig. 6(a) and (b), the cone-shaped 3D imaging space were decomposed into 60 concentric circular planes and 128 sector planes, which are perpendicular to each other. For each circular plane, a raw B-Scan image was reconstructed from the 256 A-lines (the ultrasound signals received at the 256 detection points on the circular scan path). Since each A-line has 8192 sampling points, the B-Scan image consists of a total number of 256 × 8192 pixels. For each pixel, its horizontal (lateral) location is determined by the location of the corresponding detection point, while its vertical (axial) location is determined by its travel time multiplied by the sound velocity. The grayscale or the color of the pixel represents the strength of the ultrasound signal (i.e., the recorded value at the sampling point). After this, a second B-Scan image was reconstructed based new ultrasound signal strengths determined by SAFT. Following the same procedure, for each sector plane, two B-scan images were reconstructed from the 120 A-lines (with 8192 sampling points in each) with and without applying SAFT, respectively. Next, the B-Scan images for the 60 circular planes and the 128 sector planes were normalized and combined together pixel by
––– (1),
C.-H. Huang, J. Zou / Sensors and Actuators A 236 (2015) 281–288
Fig. 6. 3D ultrasound Image reconstruction process. (a) Circular B-Scan image reconstruction. (b) Sector B-Scan image reconstruction. (c) C-Scan image slicing. (d) 3D image formation.
285
Fig. 8. Picture of the imaging target.
d = ı cos
––(9),
where d is the distance between the two c-planes bounding Ti andPj , respectively, r is the diameter of the circular path passing the detection point (Ti ), r is the diameter of the circular path passing the reconstruction pixel (Pj ), is the scanning angle of the transducer, and ϕ is the angle between the two sector planes bounding Ti and Pj , respectively. For applying SAFT on a sector plane (Fig. 7(b)), the observation distance can be calculated as =
l2 l2 − 2ll cos
(6)
Where l is the distance from the ultrasound detection point (Ti ) to the anchor point of the hinge, l is the distance from the reconstruction pixel (Pi ) to the anchor point of the hinge, and is the cross angle between Ti and Pj with respect to the anchor point. 4. Imaging experiments and results
Fig. 7. Illustration of the geometry for the calculation of the observation distance for SAFT. (a) On circular planes. (b) On sector planes.
pixel to reconstruct the 3D image reconstruction, which consists of a total number of 60 × 256 × 8192 voxels. The voxels were further regrouped onto 8192C-Scan images in Cartesian coordinates and saved into the png format (Fig. 6(c)). At last, the 8192C-Scan images were stacked together to create a 3D view in Volview (Kitware) (Fig. 6(d)). Applying 2D SAFT on the circular and sector planes is similar with that on Cartesian coordinates, except that the observation distance between a detection point and a reconstruction pixel needs to be calculated in a different way. As shown in Fig. 7(a), for applying SAFT on a circular plane, the observation distance () from an ultrasound detection point (Ti ) to a reconstruction pixel ( Pj) can be calculated as = c=
c 2 + d2
r 2 + r 2 − 2rr cos ϕ
––– (7), –– (8),
Fig. 8 shows the imaging target for the ultrasound imaging experiments with the scanning transducer. It consists of two silicon discs with a diameter of 6 mm, which are supported with four optical fibers with a diameter of 100 m. To fix the optical fibers in place, four acrylic plates with an array of positioning holes were made by laser cutting and assembled together with epoxy. The distances between the two silicon discs are 15 mm, 10 mm, and 20 mm, respectively. Due to their small diameter, the back-scattered ultrasound signals from the optical fibers will be weak and therefore would not interfere with those from the two silicon discs. During the ultrasound imaging experiment, the entire assembly was placed underneath the scanning transducer setup and completely immersed in water. 4.1. Alignment of the detection points To implement the 2D SAFT on the circular and sector planes, the ultrasound detection points need to be positioned on the 2D grid formed by the circular and sector planes. However, due to the random time delay ( ) between the trigger signals from the LabView controlled computer and the driving signals from the function generators, the ultrasound detection points on the circular scanning paths could be misaligned (Fig. 9(a)). In addition, when the scanning of the transducer is shift from one circular path to another, some ultrasound detection points could appear on the transition section between two circular paths. To address this issue, the data recording of the DAQ card was delayed by a time () (longer than ) to exclude those off-path detection points. The total recording period
286
C.-H. Huang, J. Zou / Sensors and Actuators A 236 (2015) 281–288
the circular scanning paths can be re-aligned based on the phase of the corresponding driving signals (Fig. 9(b)). 4.2. 2D image reconstruction Fig. 10(a) shows the raw B-Scan image (before applying SAFT) of a representative circular plane reconstructed from the 256 Aline data. Fig. 10(b) shows the B-Scan image of the same circular plane after applying SAFT. The reverberation signals from the ultrasound echo traveled back and forth between sample and transducer were canceled by using the reverberation nulling and echo prediction method [25–27]. The signals reflected from the aluminum plate at the bottom of acrylic holder and from the tip of transducer itself were also removed. A similar image processing procedure was adopted for the section plane. Fig. 10(c) and (d) show the B-Scan images of one representative sector plane before and after applying SAFT. Both images consist of 120 A-line data. 4.3. 3D image reconstruction
Fig. 9. (a) Random time delay in the AC driving signals and the resulting misalignment of the ultrasound detection points. (b) Re-alignment of the ultrasound detection points based on the AC driving signals.
was still equal to the period of the driving signal. The relative location of an ultrasound detection point on a circular scanning path is determined by the time and therefore the phase of the AC driving signals (see Fig. 2). Therefore, the ultrasound detections points on
Fig. 11(a) shows the reconstructed 3D image based on the raw B-Scan images of both the circular and sector planes. The C-Scan images with gradually increased diameters can be clearly seen. The margin of the graphs excluding the Cartesian coordinates is presented in white color. In order to provide a better view of the two imaging targets, most of the margin was removed by using the shrink function in Volview (Fig. 11(b)). However, in order to keep the image of the two targets intact, a small portion of the white margin at the upper-left corner has to be kept and therefore is also shown in the plot. Fig. 11(c) shows the reconstructed 3D image based on the B-Scan images of both the circular and sector planes after applying SAFT. Compared with the image without
Fig. 10. B-Scan image of a representative (flattened) circular plane (a) before and (b) after applying SAFT. B-Scan image of a representative sector plane (c) before and (d) after applying SAFT.
C.-H. Huang, J. Zou / Sensors and Actuators A 236 (2015) 281–288
287
ing target was successfully conducted. The lateral resolution of the 3D ultrasound image was improved with modified SAFT algorisms based on the unique format of the acquired ultrasound data. The scanning transducer technique requires only one DAQ channel. Therefore it does not involve the complex and expensive multi-channel DAQ electronics in the ultrasound-array based imaging systems. Ultrasound imaging at different frequency ranges can be obtained by using a single-element transducer with different frequency response. The liquid-immersible electromagnetic microactuator provided 2-axis mechanical scanning at 10s–100s Hz. This made the acquisition of all the ultrasound data for 3D image reconstruction to be performed in a timely manner. This work laid a foundation for the development of compact and low-cost hand-held scanning transducer probes for 3D ultrasound imaging. In the future, new functionalities (e.g., optical delivery) will be investigated to expand the imaging capability (e.g., photoacoustic tomography) of the scanning transducer technique.
References
Fig. 11. Reconstructed 3D ultrasound images in Volview. (a) Raw image. (b) Image with white margin removed. (c) Image with white margin removed and after applying SAFT. The red, green and blue lines indicate the x y and z axes.
SAFT, both the contrast and spatial resolution of the imaging targets were significantly improved. The field of view of the scanning transducer is a cone shape (3-cm diameter on the top, 6-cm diameter circle at the bottom, and 6-cm depth). The lateral resolution is ∼1 mm, which is limited by the diameter (2 mm) of the transducer. The mean diameters of the upper and lower silicon discs were estimated to be 6.28 mm and 5.83 mm, respectively, which close to the actual dimension (6 mm). The distances between the two silicon pieces along x, y, and z axes were estimated to be 14.2 mm, 9.88 mm, and 20.11 mm, respectively, which match well with their actual values. 5. Conclusion In this work, a new two-axis micromechanical scanning transducer technique was developed and 3D pulse-echo ultrasound imaging of two silicon discs immersed in water as the imag-
[1] P.l. Greguss, Ultrasonic Imaging: Seeing by Sound: the Principles and Widespread Applications of Image Formation by Sonic, Ultrasonic, and Other Mechanical Waves, 1st ed., Focal Press, London, New York, 1980. [2] T.L. Szabo, Diagnostic Ultrasound Imaging: Inside Out, Elsevier Academic Press, Amsterdam, Boston, 2004. [3] A. Fenster, D.B. Downey, Three-dimensional ultrasound imaging, Annu. Rev. Biomed. Eng. 2 (2000) 457–475. [4] A. Fenster, D.B. Downey, 3-D ultrasound imaging: a review, IEEE Eng. Med. Biol. Mag. 15 (November–December) (1996) 41–51. [5] J.P.S. Garcia, P.T. Garcia, R.B. Rosen, P.T. Finger, A 3-dimensional ultrasound C-scan imaging technique for optic nerve measurements, Ophthalmology 111 (June) (2004) 1238–1243. [6] Q.H. Huang, Y.P. Zheng, M.H. Lu, Z.R. Chi, Development of a portable 3D ultrasound imaging system for musculoskeletal tissues, Ultrasonics 43 (January) (2005) 153–163. [7] E.D. Light, R.E. Davidsen, J.O. Fiering, T.A. Hruschka, S.W. Smith, Progress in two-dimensional arrays for real-time volumetric imaging, Ultrason. Imaging 20 (January) (1998) 1–15. [8] G. Rizzatto, Evolution of ultrasound transducers: 1.5 and 2D arrays, Eur. Radiol. 9 (Suppl. 3) (1999) S304–S306. [9] I.O. Wygant, X. Zhuang, D.T. Yeh, O. Oralkan, A. Sanli Ergun, M. Karaman, et al., Integration of 2D CMUT arrays with front-end electronics for volumetric ultrasound imaging, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 55 (February) (2008) 327–342. [10] F. Forsberg, Ultrasonic biomedical technology; marketing versus clinical reality, Ultrasonics 42 (April) (2004) 17–27. [11] R.N. Rohling, A.H. Gee, L. Berman, Automatic registration of 3-D ultrasound images, Ultrasound Med. Biol. 24 (July) (1998) 841–854. [12] Q.H. Huang, Y.P. Zheng, R. Li, M.H. Lu, 3-D measurement of body tissues based on ultrasound images with 3-D spatial information, Ultrasound Med. Biol. 31 (December) (2005) 1607–1615. [13] V. Grau, H. Becher, J.A. Noble, Registration of multiview real-time 3-D echocardiographic sequences, IEEE Trans. Med. Imaging 26 (September) (2007) 1154–1165. [14] X. Ye, J.A. Noble, D. Atkinson, 3-D freehand echocardiography for automatic left ventricle reconstruction and analysis based on multiple acoustic windows, IEEE Trans. Med. Imaging 21 (September) (2002) 1051–1058. [15] A. Briggs, O.V. Kolosov, Acoustic Microscopy, 2nd ed., Oxford University Press, Oxford, 2010. [16] R.G. Maev, Advances in acoustic microscopy and high resolution imaging, Wiley-VCH, John Wiley distributor, Weinheim, Chichester, 2013. [17] A.S. Savoia, G. Caliano, M. Pappalardo, A CMUT probe for medical ultrasonography: from microfabrication to system integration, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 59 (June) (2012) 1127–1138. [18] M.T. Myaing, D.J. MacDonald, X.D. Li, Fiber-optic scanning two-photon fluorescence endoscope, Opt. Lett. 31 (April) (2006) 1076–1078. [19] C.M. Lee, C.J. Engelbrecht, T.D. Soper, F. Helmchen, E.J. Seibel, Scanning fiber endoscopy with highly flexible, 1mm catheterscopes for wide-field, full-color imaging, J. Biophotonics 3 (June) (2010) 385–407. [20] G.J. Tearney, S.A. Boppart, B.E. Bouma, M.E. Brezinski, N.J. Weissman, J.F. Southern, et al., Scanning single-mode fiber optic catheter-endoscope for optical coherence tomography, Opt. Lett. 21 (April) (1996) 543–545. [21] E.J. Seibel, Q.Y.J. Smithwick, C.M. Brown, P.G. Reinhall, Single fiber flexible endoscope: general design for small size, high resolution, and wide field of view, Biomonitoring Endosc. Technol. 1 (2001) 29–39. [22] Z.L. Deng, X.Q. Yang, H. Gong, Q.M. Luo, Two-dimensional synthetic-aperture focusing technique in photoacoustic microscopy, J. Appl. Phys. 109 (May) (2011).
288
C.-H. Huang, J. Zou / Sensors and Actuators A 236 (2015) 281–288
[23] C.H. Frazier, W.D. O’Brien, Synthetic aperture techniques with a virtual source element, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 45 (January) (1998) 196–207. [24] N.V. Ruiter, G.F. Schwarzenberg, M. Zapf, H. Gemmeke, Conclusions from an Experimental 3D Ultrasound Computer Tomograph, 2008 IEEE Nuclear Science Symposium and Medical Imaging Conference (2008 Nss/Mic), vols. 1–9, pp. 3776–3783, (2009). [25] T. Folegot, C. Prada, M. Fink, Resolution enhancement and separation of reverberation from target echo with the time reversal operator decomposition, J. Acoust. Soc. Am. 113 (June) (2003) 3155–3160. [26] S.G. Parulekar, Ultrasonic detection of calcification in gallstones: the reverberation shadow, J. Ultrasound Med. 3 (March) (1984) 123–129. [27] H.C. Song, W.S. Hodgkiss, W.A. Kuperman, K.G. Sabra, T. Akal, M. Stevenson, Passive reverberation nulling for target enhancement, J. Acoust. Soc. Am. 122 (December) (2007) 3296–3303.
Biographies
Chih-Hsien HuangChih-Hsien Huang received his B.S. and M.S. degrees in Electrical Engineering from the National Cheng Kung University, Tainan, Taiwan, in 2003 and 2005, respectively. From 2005 to 2010, he was with Industrial Technology Research Institute—A Research and Development Foundation, Tainan, Taiwan, where he developed various feedback control methodologies for LED backlight system including current mixing architecture, flick free sequential detection method, and wavelength stabling algorithm. He is currently a Ph.D. student in the Department of Electrical and Computer Engineering at Texas A&M University. His research interest is focused on MEMS based scanning devices for acoustic tomography applications.
Jun Zou received his Ph.D. degree in electrical engineering from the University of Illinois at Urbana-Champaign in 2002. In 2004, he joined in the department of electrical and computer engineering at Texas A&M University, where he is currently an associate professor. His current research interests lie in the development of micro and nano optoelectro-mechanical devices and systems for biomedical imaging and sensing applications.