Image Quality Evaluation with a New Phase Rotation Beamformer

Transcription

1 IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, vol. 55, no. 9, September Image Quality Evaluation with a New Phase Rotation Beamformer Anup Agarwal, Student Member, IEEE, Fabio Kurt Schneider, Member, IEEE, Yang Mo Yoo, Member, IEEE, and Yongmin Kim, Fellow, IEEE Abstract Over the last few decades, dynamic focusing based on digital receive beamforming (DRBF) has led to significant improvements in image quality. However, it is computationally very demanding due to its requirement for multiple lowpass filters (e.g., a complex filter for each receive channel in quadrature demodulation-based phase rotation beamformers (QD-PRBF)). We recently developed a novel phase rotation beamformer with reduced complexity, which can lower: 1) the number of lowpass filters using 2-stage demodulation (TSD) and 2) the number of beamforming points using adaptive field-of-view (AFOV) imaging. In TSD, dynamic focusing is performed on the mixed signals, while sampling frequency of the beamformed signal (i.e., beamforming frequency) is adjusted based on the displayed field-of-view (FOV) size in AFOV imaging. In this paper, the image quality of the developed beamformer (i.e., TSD-AFOV-PRBF) has been quantitatively evaluated using phantom and in vivo data. From the phantom study, it was found that TSD-AFOV-PRBF with only 1024 beamforming points provides comparable image quality to QD-PRBF. We obtained a median contrast resolution (CR) degradation of 7.6% for the FOV size of 160 mm. Image quality steadily improves with FOV size reduction (e.g., 2.3% CR degradation at 85 mm). Similar results were also obtained from an in vivo study. Thus, TSD-AFOV-PRBF could provide comparable image quality to conventional beamformers at considerably reduced computational cost. I. Introduction The widespread use of digital receive beamforming (DRBF) based on interpolation or phase rotation has contributed greatly to improved image quality in modern ultrasound systems [1]. However, both interpolation and quadrature-demodulation-based phase rotation beamformers (IBF and QD-PRBF) are computationally very expensive due to their requirement of a finite impulse response (FIR) filter for each receive channel [2] [4]. Toward reducing the complexity of DRBF, several methods have been proposed [5] [10]. Although these methods can lower the computational requirement for receive beamforming, they tend to introduce substantial degradation in image quality [11]. Manuscript received September 21, 2007; accepted February 23, A. Agarwal, Y. M. Yoo, and Y. Kim are with Image Computing Systems Laboratory Departments of Electrical Engineering and Bioengineering University of Washington, Seattle, WA ( ykim@ u.washington.edu). F. K. Schneider is with Laboratory of Image Computing and Electronic Instrumentation Grad School of Electrical Engineering and Applied Computer Science Federal University of Technology, Paraná, Brazil. Digital Object Identifier /TUFFC.886 With the goal of reducing the computational complexity without incurring much degradation in image quality, we previously developed a novel phase rotation beamformer [11] [13]. This beamformer performs dynamic focusing on the mixed signals, i.e., 2-stage demodulation (TSD) [11] and adjusts the beamforming frequency based on the displayed field-of-view (FOV) characteristics, i.e., adaptive FOV (AFOV) imaging [12], [13]. It was found that the developed TSD-AFOV-PRBF greatly reduces the computational requirement of a B-mode imaging system (e.g., 86.8% reduction in number of operations compared with QD-PRBF when 1024 beamforming points are used) [13]. In this paper, we have extensively evaluated the image quality of the developed beamformer for various FOV sizes using phantom and in vivo data. II. Methods and Materials A. New Low-Cost Phase Rotation Beamformer Fig. 1 shows the block diagram of the newly developed receive beamformer called TSD-AFOV-PRBF. In TSD-AFOV-PRBF, dynamic focusing is performed on the mixed signals, while the computationally demanding demodulation filtering is performed after phase rotation and summation (i.e., 2-stage demodulation) [11]. Thus, the number of lowpass filters is substantially reduced compared with QD-PRBF. On the other hand, artifacts could be introduced due to: 1) nonlinearity in dynamic receive focusing and 2) signal aliasing at lower beamforming frequencies (f bf ) due to the presence of signal replicas at ± 2f0. It should be noted that fbf refers to the sampling frequency of the beamformed signal, while the delay quantization for supporting receive beamforming is determined by the ADC sampling frequency (fadc). Nonlinearity in dynamic focusing was found to have a negligible impact on image quality [11]. On the other hand, signal aliasing at reduced beamforming frequencies can lead to image quality degradation. Using bandpass sampling principles, we found that fbf 3f0 is required to avoid signal aliasing for BW f 0, which is typical in B- mode ultrasound imaging. Furthermore, for beamforming frequencies between 1.33f0 and 2.66f0, it was found that fbf of 1.33f0 : 1) provides the minimum signal aliasing and 2) does not suffer from signal aliasing for BW < 0.66f0 [11], [12] /$ IEEE

2 1948 IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, vol. 55, no. 9, September 2008 TABLE I. Parameters Used for Phantom Data Acquisition. Parameter Value Number of channels (transmit and receive) 32 Speed of sound (m/s) 1540 Transducer, center frequency (MHz) Linear, 4.5 MHz Convex, 3.5 MHz Receive signal bandwidth ( 45 db) 1.4f 0 Sampling frequency (MHz) 40 MHz Transmit focus (mm) Varied Number of scanlines 128 Fig. 1. Block diagram of the proposed TSD-AFOV-PRBF. In TSD-AFOV-PRBF, an adaptive FOV controller adjusts the beamforming frequency based on: 1) the displayed FOV properties (i.e., axial size and mean center frequency) and 2) the number of available beamforming points per scanline. Because only the displayed area is reconstructed in AFOV imaging, high spatial resolution can be achieved for small field-of-view sizes even with a reduced number of beamforming points. It should be noted that ADC sampling frequency is same for all the FOVs. Thus, the delay quantization does not vary with FOV size. On the other hand, the maximum beamforming frequency in AFOV imaging is given by c L f bf = 2 D where c is the speed of sound in the medium, D is the axial FOV size, and L is the number of beamforming points. Although (1) determines the maximum beamforming frequency, the center frequency of received signals needs to be estimated before selecting fbf due to the fact that the ratio of fbf to f0 has a large impact on signal aliasing in TSD-PRBF. It was found that the developed beamformer with 1024 beamforming points decreases the overall computing requirement by 86.8% (in terms of number of operations) compared with quadrature-demodulation-based phase rotation beamforming [13]. In this paper, we have extensively evaluated the image quality of TSD-AFOV-PRBF with 1024 beamforming points. B. Experimental Setup 1. Phantom Data Acquisition Setup: Pre-beamformed data were acquired from a Sonix RP scanner (Ultrasonix Medical Corporation, Vancouver, Canada) using the Texo research interface [14], [15]. The data was acquired one channel at a time. Two tissue-mimicking phantoms (i.e., RMI 403 GS and RMI 404 GS, Gammex RMI, Middleton, WI) and one abdominal phantom (Computerized Imaging Reference Systems Model 57, Norfolk, VA) were used. Various acquisition parameters in Table I were used. The transmit focus was selected based on the target location for each phantom and was not modified for different FOV sizes. For example, a transmit focus of 30 mm was selected (1) while visualizing the point target in RMI 403 GS; the reconstructed image is shown in Fig. 2(a). Acquired pre-beamformed data were processed off-line to compare the performance of the newly developed beamformer to that of QD-PRBF. In QD-PRBF, a beamforming frequency of 18 MHz was used for a 4.5-MHz linear array transducer, while a beamforming frequency of 14 MHz was used for a 3.5-MHz convex array transducer. On the other hand, the beamforming frequency in TSD- AFOV-PRBF was adjusted depending on the field-of-view and received signal s center frequency. Table II shows the beamforming frequencies for linear and convex array transducers change as the FOV is adjusted. The maximum beamforming frequency was limited to 4f0. 2. In Vivo Data: Pre-beamformed in vivo data obtained from healthy volunteers using an SA-9900 scanner (Medison, Seoul, Korea) were used. A 192-element convex array transducer with a center frequency of 3.5 MHz was used. A 64-element aperture was used for transmission, while 32 channels were used for receiving the ultrasound signals. The transmit focus was adjusted based on the depth of the region of interest. The received signals were quantized using 8-bit ADCs running at MHz. 3. Evaluation Metrics: For quantitative comparison, spatial resolution and contrast resolution (CR) were computed for various FOV sizes with the phantom data, while CR was computed with the in vivo data. For spatial resolution, we evaluated the 20-dB axial and lateral resolution from the point targets by measuring the lateral and axial beamwidths at the 20dB level (the point spread function was interpolated before the measurement). On the other hand, CR was measured by I t - I s CR = st + s (2) 2 2 s where It and Is are the average intensity in the target and speckle regions, respectively, and s t 2 and s s 2 are the variance of each region. A. Phantom Study III. Results and Discussion 1. Tissue Mimicking Phantoms: Fig. 2(a) and 2(b) show 2 example ultrasound images from the RMI 403 GS phan-

3 Agarwal et al.: image quality evaluation with a new phase rotation beamformer 1949 tom obtained using the 4.5-MHz linear array transducer and reconstructed with QD-PRBF and FOV of 55 mm. Fig. 2(c) shows the 20 db axial beamwidth for the point target circled in Fig. 2(a). The developed beamformer shows some degradation in axial resolution for FOV of 160 mm (i.e., 20-dB axial resolution of 1.05 mm and 0.82 mm with TSD-AFOV-PRBF and QD-PRBF, respectively). Reducing the FOV size improves the axial resolution with TSD-AFOV-PRBF (i.e., axial resolution of 0.89 mm and 0.82 mm for FOV of 112 mm and 56 mm, respectively). Similarly, the developed beamformer shows less than 1% degradation in lateral resolution for all FOVs except 160 mm as shown in Fig. 2(d). Fig. 2(e) shows CR obtained for various FOV sizes. Consistent with the axial resolution results, TSD-AFOV-PRBF shows comparable CR for all FOV sizes except 160 mm. Consistent results were also obtained with the other tissue mimicking phantom (i.e., RMI 404 GS). 2. Abdominal Phantom: Multiple data sets were obtained from the abdominal phantom by varying the scanning location and transducer angle of the 3.5-MHz convex array transducer. Fig. 3 shows the B-mode images reconstructed using QD-PRBF as well as the cysts and targets used for the contrast resolution evaluation. Table III lists the CR values measured for these cysts and targets using the TSD-AFOV-PRBF. The entries of the table marked with an asterisk represent cases where the CR value from TSD-AFOV-PRBF is worse by more than 5% compared with QD-PRBF. As shown in Table III, the developed beamformer provides less than 5% degradation for FOVs of 85 mm, 56 mm, and 28 mm. For larger FOVs (i.e., 160 mm and 112 mm), it yields CR degradation of greater than 5% for 6 out of the 8 data sets evaluated. The greatest CR degradation of 12.2% was obtained for target 0 in data set 2 when displayed with a FOV of 160 mm. To assess this CR degradation, Fig. 4 shows the B-mode images reconstructed using the 2 beamforming methods. It is difficult to differentiate between the ultrasound images obtained from the 2 methods. Thus, despite yielding some CR degradation, TSD-AFOV-PRBF can provide visually comparable image quality to QD-PRBF for larger FOV sizes. For further analysis, Fig. 5 shows the boxplot of percent CR degradation with TSD-AFOV-PRBF for all 15 targets in the 7 images. For FOV of 160 mm, median CR degradation with the developed beamformer is 7.6%, which does not lead to noticeable image quality degradation. As the FOV is reduced, the image quality of TSD-AFOV-PRBF steadily improves. For example, for FOV of 85 mm, median degradation is 2.3%. Similarly, further size reduction to 56 mm increases the beamforming frequency to 4f0, which results in no degradation compared with QD-PRBF. B. In Vivo Evaluation Fig. 2. Phantom study results with the RMI 403 GS phantom: B-mode images with QD-PRBF showing (a) point targets, (b) cysts, (c) 20-dB axial resolution, (d) 20-dB lateral resolution, and (e) contrast resolution with QD-PRBF and TSD-AFOV-PRBF. The in vivo data sets were reconstructed using QD- PRBF and TSD-AFOV-PRBF, and the contrast resolutions were measured in a similar manner. Fig. 6 shows the various in vivo data sets reconstructed using QD-PRBF as well as the cysts and targets used for contrast evaluation. Table IV lists the CR values obtained using the 2 beamforming methods, in which the entries marked with an asterisk represent cases where the CR value from TSD- AFOV-PRBF is worse by more than 5% compared with

4 1950 IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, vol. 55, no. 9, September 2008 Fig. 3. Abdominal phantom images reconstructed using QD-PRBF: (a) data set 0, (b) data set 1, (c) data set 2, (d) data set 3, (e) data set 4, (f) data set 5, (g) data set 6, and (h) data set 7. the CR value obtained using the QD-PRBF. Similar to the phantom study results, there is small CR degradation for larger FOVs (i.e., 112 mm and 160 mm). For example, contrast resolutions of 3.36 and 3.11 were measured from QD-PRBF and TSD-AFOV-PRBF, respectively, with cyst 0 in data set 1 for FOV of 160 mm. Fig. 7 shows the reconstructed ultrasound images. We can observe in Fig. 7 that there is no substantial negative impact of this slightly degraded CR on image quality. As the FOV size reduces, the beamforming frequency for the developed beamformer increases resulting in improved CR performance. For example, the CR value for the same cyst improves from 3.11 (for 160 mm FOV) to 3.28 and 3.36 for FOVs of 85 mm and 56 mm, respectively. Fig. 8 shows the boxplot of percent CR degradation with the developed beamformer. TSD-AFOV-PRBF yields median CR degradation less than 10% for all evaluated FOVs. For example, median CR degradation of 7.9% was obtained for FOV of 160 mm as shown in Fig. 8. Similar CR values were obtained for the 112 mm and 160 mm FOVs, because the same beamforming frequency (i.e., 1.33f0) was used for both FOVs. Reducing the FOV to 85 mm significantly improves the CR performance, resulting in median degradation of 2.4%. For FOV of 56 mm and smaller, the beamforming frequency increases to 4f0, resulting in no CR degradation compared with the conventional beamformer. Thus, from the phantom and in vivo studies, it was found that TSD-AFOV-PRBF does not yield substantial CR degradation. Although it incurs small resolution degradation for larger FOVs, it might not be perceptible clinically (based on the preliminary phantom and in vivo evaluations). Hence, with 1024 beamforming points, TSD- TABLE II. FOV Sizes and Their Corresponding Beamforming Frequencies for TSD-AFOV-PRBF when F 0 = 3.5 MHz And 4.5 MHz and the Number of Beamforming Points is FOV size 3.5 MHz 4.5 MHz 160 mm 1.33f0 1.13f0 112 mm 1.33f0 1.33f0 85 mm 2.90f0 1.33f0 56 mm 4f0 3.23f0 28 mm 4f0 4f0

5 Agarwal et al.: image quality evaluation with a new phase rotation beamformer 1951 Fig. 4. Abdominal phantom images for FOV of 160 mm reconstructed using (a) QD-PRBF and (b) TSD-AFOV-PRBF. AFOV-PRBF provides comparable image quality to QD- PRBF while substantially reducing the hardware requirements. C. TSD-AFOV-PRBF with 512 Beamforming Points By reducing the number of beamforming points to 512, even greater computational savings can be achieved (i.e., 48.1% reduction in the number of operations compared with using 1024 points for a 32-channel B-mode system [13]). Fig. 9 shows ultrasound images obtained with the abdominal phantom when the number of beamforming points is reduced to 512. As shown in Fig. 9(a), there is definite image quality degradation for FOV of 160 mm compared with QD-PRBF as seen in Fig. 4(a). On the other hand, as the FOV is reduced, there is an improvement in image quality as shown in Fig. 9(b) for FOV of 85 mm. Thus, with 512 beamforming points, the developed beamformer could provide comparable image quality for smaller FOVs while yielding perceptible degradation for larger FOVs. The performance was further evaluated by measuring CR with the phantom and in vivo data sets. Data set 0 Data set 1 Data set 2 Data set 3 TABLE III. Phantom Results: CR Values Obtained for Various Data Sets With QD-PRBF and TSD-AFOV-PRBF. Target 1 Target 2 Data set 4 Data set 5 Target 1 Data set 6 Data set mm 112 mm 85 mm 56 mm 28 mm QD-PRBF TSD-AFOV-PRBF QD-PRBF TSD-AFOV-PRBF QD-PRBF TSD-AFOV-PRBF 1.98* 1.98* QD-PRBF TSD-AFOV-PRBF 1.53* 1.53* QD-PRBF TSD-AFOV-PRBF 0.72* 0.72* QD-PRBF TSD-AFOV-PRBF QD-PRBF TSD-AFOV-PRBF 1.89* 1.89* QD-PRBF TSD-AFOV-PRBF 0.39* 0.39* QD-PRBF TSD-AFOV-PRBF 2.05* 2.05* QD-PRBF TSD-AFOV-PRBF 2.27* 2.27* QD-PRBF TSD-AFOV-PRBF 2.36* 2.36* QD-PRBF TSD-AFOV-PRBF 2.14* 2.14* QD-PRBF TSD-AFOV-PRBF 3.49* 3.49* QD-PRBF TSD-AFOV-PRBF QD-PRBF TSD-AFOV-PRBF * Asterisks denote CR values from TSD-AFOV-PRBF that are more than 5% worse than those from QD-PRBF.

6 1952 IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, vol. 55, no. 9, September 2008 Fig. 5. Percent degradation in CR with TSD-AFOV-PRBF in the abdominal phantom study. Fig. 10 shows the boxplot of CR degradation compared with QD-PRBF for all targets and cysts in the phantom and in vivo data sets. Consistent with the visual image quality results, there is substantial CR degradation for larger FOVs with 512 beamforming points. For example, when the FOV is 160 mm, we can see median CR degradation of 32.8% and 26.6% in Fig. 10(a) and 10(b), respectively. A small CR improvement is observed with FOV of 112 mm. However, it is not statistically significant because there is a large overlap between the boxplots from 112 mm and 160 mm for both phantom and in vivo data as shown in Fig. 10(a) and 10(b). On the other hand, reducing FOV to 85 mm significantly lowers CR degradation, resulting in comparable image quality to QD-PRBF. In this paper, we have conducted an image quality evaluation of a newly developed phase rotation beamformer in its default setting (i.e., 1024 beamforming points) as well as with a reduced number of points (i.e., 512). It is found that the number of beamforming points in TSD-AFOV- PRBF substantially influences the image quality obtained. For example, in its default setting, image quality similar Fig. 6. In vivo images reconstructed using QD-PRBF: (a) data set 0, (b) data set 1, (c) data set 2, (d) data set 3, (e) data set 4, (f) data set 5, (g) data set 6, and (h) data set 7.

7 Agarwal et al.: image quality evaluation with a new phase rotation beamformer 1953 Fig. 7. In vivo images for FOV of 160 mm reconstructed using (a) QD-PRBF and (b) TSD-AFOV-PRBF. to conventional beamformers can be obtained. This would be advantageous for development of ultrasound systems, in which there is a need for hardware savings without substantial image quality degradation (e.g., next-generation 3-D ultrasound systems with thousands of transducer elements). Furthermore, the computational savings offered by the developed beamformer (with 1024 or 512 points) can facilitate development of smaller, more compact portable ultrasound systems by enabling single-chip solutions for ultrasound processing (i.e., a single chip to support both front- and back-end processing requirements in medical ultrasound systems). We believe that these single-chip solutions would eventually lead to integration of signal acquisition and processing stages (i.e., processing inside the scanhead), thereby eliminating the expensive highbandwidth, low-noise connections from the transducer to the ultrasound processing unit. Finally, our developed beamformer would be beneficial for programmable computing architectures, which have emerged as a viable alternative to application-specific integrated circuits for supporting ultrasound signal processing [16] [19]. It has been recently shown that a hybrid programmable architecture consisting of a low-cost field programmable gate array and a digital signal processor (DSP) can support all the front- and back-end processing in low-end ultrasound systems [18], [19]. We believe that the savings offered by the developed beamformer could facilitate compact programmable architectures (e.g., single chip solution based on a low-cost DSP). In addition, by exploiting the flexibility offered by programmable proces- Data set 0 Data set 1 Data set 2 TABLE IV. In Vivo Results: CR Values Obtained for Various Data Sets with QD-PRBF and TSD-AFOV-PRBF. Cyst 2 Data set 3 Data set 4 Data set 5 Data set 6 Data set mm 112 mm 85 mm 56 mm 28 mm QD-PRBF TSD-AFOV-PRBF QD-PRBF TSD-AFOV-PRBF 1.53* 1.53* QD-PRBF TSD-AFOV-PRBF 3.11* 3.11* QD-PRBF * TSD-AFOV-PRBF 1.53* 1.53* QD-PRBF TSD-AFOV-PRBF QD-PRBF TSD-AFOV-PRBF QD-PRBF TSD-AFOV-PRBF 1.69* 1.69* QD-PRBF TSD-AFOV-PRBF QD-PRBF TSD-AFOV-PRBF 3.24* 3.24* QD-PRBF TSD-AFOV-PRBF 1.23* 1.23* QD-PRBF TSD-AFOV-PRBF 1.27* 1.27* QD-PRBF TSD-AFOV-PRBF 1.57* 1.57* QD-PRBF TSD-AFOV-PRBF 2.36* 2.36* * Asterisks denote CR values from TSD-AFOV-PRBF that are more than 5% worse than those from QD-PRBF.

8 1954 IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, vol. 55, no. 9, September 2008 could be used, resulting in significantly reduced front-end processing requirements. The remaining computing resources can be used to support imaging analysis and/or image segmentation. On the other hand, for applications where high-resolution imaging is needed for larger FOVs, the number of beamforming points in TSD-AFOV-PRBF can be increased (e.g., 1024 points) by allocating more computing resources for front-end processing. IV. Conclusions Fig. 8. Percentage degradation in CR with TSD-AFOV-PRBF in the in vivo study. sors (e.g., DSPs), the same architecture could be used to support the requirements of different applications by dynamically adjusting the number of beamforming points in TSD-AFOV-PRBF. For example, for applications where a low-resolution display for large FOVs might be acceptable, a smaller number of beamforming points (e.g., 512) In this paper, we have evaluated image quality with the newly developed phase rotation beamformer (i.e., TSD- AFOV-PRBF). From phantom and in vivo studies, it was found that the developed beamformer incurs small image quality degradation for larger FOV sizes. As the FOV decreases, the image quality significantly improves, resulting in a negligible difference compared with conventional beamformers. Even greater computational savings can be achieved by reducing the number of beamforming points (e.g., 512) at the cost of degraded image quality for large FOV sizes. We believe that TSD-AFOV-PRBF would be beneficial in the development of 3-D ultrasound systems Fig. 9. Ultrasound images reconstructed using TSD-AFOV-PRBF with 512 beamforming points for FOV of (a) 160 mm and (b) 85 mm. Fig. 10. Percent degradation in CR with TSD-AFOV-PRBF when the number of beamforming points is 512 with (a) phantom data and (b) in vivo data.

9 Agarwal et al.: image quality evaluation with a new phase rotation beamformer 1955 with thousands of elements as well as ultraportable handheld ultrasound systems. Acknowledgments We would like to thank Dr. Shahram Vaezy at the University of Washington for allowing us to use the Ultrasonix RP for phantom data acquisition and Dr. T. K. Song from Sogang University (Seoul, Korea) for providing us with the in vivo data sets. References [1] S. Stergiopoulos, Advanced Signal Processing Handbook: Theory and Implementation for Radar, Sonar, and Medical Imaging Real Time Systems. Boca Raton, FL: CRC Press, 2000, pp [2] B. D. Steinberg, Digital beamforming in ultrasound, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 39, pp , [3] J. N. Wright, C. R. Cole, and A. Gee, Method and apparatus for a baseband processor for a receive beamformer system, U.S. Patent No , [4] M. O Donnell, W. E. Engeler, J. J. Bloomer, and J. T. Pedicone, Method and apparatus for digital phase array imaging, U.S. Patent No , [5] K. Ranganathan, M. K. Santy, T. N. Blalock, J. A. Hossack, and W. F. Walker, Direct sampled I/Q beamforming for compact and very low-cost ultrasound imaging, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 51, pp , [6] S. R. Freeman, M. K. Quick, M. A. Morin, R. C. Anderson, C. S. Desilets, T. E. Linnenbrink, and M. O Donnell, Delta-sigma oversampled ultrasound beamformer with dynamic delays, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 46, pp , [7] M. Kozak and M. Karaman, Digital phased array beamforming using single-bit delta-sigma conversion with non-uniform oversampling, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 48, pp , [8] S. E. Noujaim, S. L. Garverick, and M. O Donnell, Phased array ultrasonic beam forming using oversampled A/D converters, U.S. Patent No , [9] S. H. Chang, K. I. Jeon, M. H. Bae, and S. B. Park, Samplingpoint-wise digital dynamic focusing in the receive mode of an ultrasonic linear array imaging system, in Proc. IEEE Symp. Ultrasonics, 1993, pp [10] J. H. Kim, T. K. Song, and S. B. Park, A pipelined sampled delay focusing in ultrasound imaging systems, Ultrason. Imaging, vol. 9, pp , [11] A. Agarwal, Y. M. Yoo, F. K. Schneider, C. Gao, L. M. Koh, and Y. Kim, New demodulation method for efficient phase rotation-based beamforming, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 54, pp , [12] A. Agarwal, F. K. Schneider, Y. M. Yoo, and Y. Kim, Adaptive field-of-view imaging for efficient phase rotation beamforming, in Proc. IEEE Ultrasonics Symp., 2006, pp [13] A. Agarwal, Y. M. Yoo, F. K. Schneider, and Y. Kim, Adaptive field-of-view imaging for efficient receive beamforming in medical ultrasound systems, Ultrasonics, in press. [14] [15] T. Wilson, J. Zagzebski, T. Varghese, Q. Chen, and M. Rao, The Ultrasonix 500RP: A commercial ultrasound research interface, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 53, pp , [16] S. Sikdar, R. Managuli, L. Gong, V. Shamdasani, T. Mitake, T. Hayashi, and Y. Kim, A single mediaprocessor-based programmable ultrasound system, IEEE Trans. Inf. Technol. Biomed., vol. 7, pp , [17] V. Shamdasani, R. Managuli, S. Sikdar, and Y. Kim, Ultrasound color-flow imaging on a programmable system, IEEE Trans. Inf. Technol. Biomed., vol. 8, pp , [18] F. K. Schneider, Fully programmable computing architecture for medical ultrasound machines, Ph.D. dissertation, Univ. Washington, Seattle, [19] T. Fukuoka, F. K. Schneider, Y. M. Yoo, A. Agarwal, and Y. Kim, Ultrasound color Doppler imaging on a fully programmable architecture, in Proc. IEEE Ultrasonics Symp., 2006, pp Anup Agarwal received the B.Tech. (Hons.) degree in electronics and electrical communications from the Indian Institute of Technology, Kharagpur, India, in 2000 and the M.S. and Ph.D degrees in electrical engineering from the University of Washington, Seattle, WA. His Ph.D. research was on low-cost receive beamforming for enabling ultra-portable handheld ultrasound systems. He is currently working at Philips Healthcare as a Systems Design Engineer. His research interests include both low-end and high-end beamforming, Doppler ultrasound, and technologies for compact ultrasound. Fabio Kurt Schneider received the B.S. in Electrical Engineering and the M.Sc. in Biomedical Engineering in 1989 and 1995, respectively, both from Federal Technological University of Parana, Parana, Brazil. He received his Ph.D. from the University of Washington in Electrical Engineering in Since 1995, he has been with the Federal Technological University of Parana in the Academic Department of Electronics. In addition to teaching and researching, he has been involved in development activities with industry since His research interests have been in areas including biomedical signal and image processing, ultrasound imaging, bioinstrumentation, digital systems design based on high-performance digital signal processors, reconfigurable devices, and ASICs. Yang Mo Yoo received the Bachelor and Master of Science degrees of Electronic Engineering in 1999 and 2001, respectively, from the Sogang University, Seoul, Korea. He received the Ph.D. degree in Bioengineering from the University of Washington, Seattle, WA. He is currently working at Philips Healthcare, Bothell, WA. Dr. Yoo has been researching and developing a hand-held ultrasound machine that would allow clinicians to have access to patient remotely and be able to quickly make diagnostic decisions. Also, he has been working for developing real-time 3-D color Doppler imaging where 3-D anatomic structures are integrated with flow information. His research interests include digital signal processing, system design, and high-performance computing architecture for medical imaging and its clinical applications. Yongmin Kim (S 79 M 82 SM 87 F 96) received the B.S. degree in electronics engineering from Seoul National University, Seoul, Korea, and the M.S. and Ph.D. degrees in electrical engineering from the University of Wisconsin, Madison, WI. He is Professor of Bioengineering and Electrical Engineering and Adjunct Professor of Radiology and of Computer Science and Engineering at the University of Washington, Seattle, WA. His research interests are in distributed diagnosis and home healthcare, medical imaging, image processing and analysis, and high-performance processor architecture. He has more than 450 publications, and his group has 72 patents and 25 commercial licenses. Dr. Kim received the Early Career Achievement Award from the IEEE Engineering in Medicine and Biology Society (EMBS) in 1988 and the 2003 Ho-Am Prize in Engineering. He is a Member of the Steering Committee for the IEEE Transactions on Medical Imaging and the Editorial Board for the IEEE Transactions on Biomedical Engineering, and IEEE Transactions on Information Technology in Biomedicine. He was the President of the IEEE EMBS in 2005 and 2006.