Real-time photoacoustic (PA) imaging involves beamforming methods using an assumed fixed sound speed, typically 1540 m/s in soft tissue. This leads to degradation of PA image quality because the true sound speed changes as PA signal propagates through different types of soft tissues: the range from 1450 m/s to 1600 m/s. This paper proposes a new method for estimating an optimal sound speed to enhance the cross-sectional PA image quality. The optimal sound speed is determined when coherent factor with the sound speed is maximized. The proposed method was validated through simulation and ex vivo experiments with microcalcification-contained breast cancer specimen. The experimental results demonstrated that the best lateral resolution of PA images of microcalcifications can be achieved when the optimal sound speed is utilized.
© 2012 OSA
Photoacoustic (PA) imaging is an emerging technology that takes advantages of an optical contrast and ultrasonic resolution and provides functional information such as hemoglobin concentration and oxygen saturation under some conditions [1–3]. The acoustic waves (i.e., PA signals) are generated only inside the targets absorbing incident laser pulses due to thermoelastic expansion. The PA signals are detected by an ultrasound (US) transducer. Depending on the detection scheme, various image reconstruction techniques can be applied such as PA microscopy (PAM), PA tomography (PAT), and cross-sectional PA imaging [3–5]. Among them, the cross-sectional PA imaging is capable of real-time imaging after minimal modification of a current commercial US scanner [5, 6]. Furthermore, this is a dual-modality imaging method combining PA and US images, thus simultaneously providing anatomical and functional information, both of which are important in the diagnosis of disease.
The real-time PA imaging involves beamformation in which the received PA signals are coherently summed by compensating for their different traveling distances (or phase differences) to reconstruct an image of initial pressure distribution. Delay-and-sum beamforming (DAS-BF) and adaptive beamformings based on coherence factor (CF) and minimum variance (MV) have been used [7, 8]. These beamforming algorithms generally assume a fixed sound speed (e.g., 1540 m/s) in calculating time delays for the receive focusing. However, the true sound speed changes as PA signal propagates through different types of soft tissues: the range from 1450 m/s in fat to 1600 m/s in muscle. The disparity between an assumed constant sound speed and the true value leads to defocusing and consequently degradation of image quality in both US and PA imaging . To mitigate the problem in PA imaging, the sound speed estimation methods have been proposed, which are suitable for PAM and PAT [10, 11]. In these methods, the sound speed in a propagating medium was estimated by finding the maximal sharpness of prominent features in image domain with gradient based methods or by fitting the analytic expression of cross-correlation with received radio-frequency (RF) channel data from an annular array transducer. The gradient based method employed a Fourier-transform-based reconstruction algorithm that less suffers from image artifacts (e.g., sidelobe) than the beamforming algorithms . However, the gradient based method is inappropriate for cross-sectional PA images reconstructed by the beamforming algorithms. This is so because the beamformation with a wrong sound speed leads to severely increasing sidelobe levels that are regarded as edges (or high sharpness) in the gradient based method. Also, the slopes of mainlobe and sidelobes vary depending on imaging depth due to the diffraction of ultrasound and the change in slice thickness. These characteristics of the cross-sectional PA imaging may cause degrading the performance of the gradient based method.
This paper proposes a new method for estimating an optimal sound speed to enhance the cross-sectional PA image quality, e.g., lateral resolution. In the method, the optimal sound speed is determined when a CF value with the sound speed is maximized. Note that the CF value is utilized not only as a focusing quality factor in estimating the optimal sound speed, but also for the adaptive beamforming based on CF to suppress the sidelobe levels [7, 8]. In addition, the estimated sound speed is not the true value but the one providing the highest spatial resolution in the region of interest (ROI). To ascertain the performances of the proposed method, ex vivo experiments were conducted with the microcalcification-contained breast cancer specimen of a volunteer patient.
2.1 Optimal sound speed estimation
As shown in Fig. 1(a) , PA signals generated from a point-like absorber are propagated toward the US array transducer and received by each element at different time. In DAS-BF, the receive focusing is dynamically conducted to form an image. To focus the PA signals at an imaging point at , the time delay for the nth element located at is calculated byEq. (1), the time delays for the dynamic receive focusing are a function of the sound speed and applied to compensate for phase differences of PA signals received at each channel. If the sound speed is incorrectly assumed, therefore, the PA signals at each channel will have still phase differences after the compensation as shown in Fig. 1(b).
CF values are computed with delay-compensated RF data by usingEq. (2), CF values are determined by coherent energy over incoherent energy of delayed RF data, which is also a function of the sound speed. When the assumed sound speed is equal to the true value of the propagation medium, the CF corresponding to the main lobe becomes the maximum (i.e., unity), while the CF for sidelobes is small. On the other hands, if the assumed sound speed is different from the actual one, the broadened beam is inevitable due to the defocusing, thus leading to decreasing the CF for the main lobe and increasing the CF for sidelobes. Consequently, the optimal sound speed in ROI can be determined when the CF related to the main lobe in the ROI is maximized with the sound speed. This can be expressed as follows:
In the proposed estimation, the pre-beamformed RF data in ROI are first captured. With an initial sound speed, the time delays for receive focusing are calculated using Eq. (1). The time delays are used to compensate for the phase differences of the captured RF data, which are also used in computing the CF in Eq. (2). As indicated in Eq. (3), the CF values are iteratively computed with the RF data in the ROI while changing the sound speed. Finally, the sound speed providing the maximal CF is determined as an optimal sound speed.
2.2 Experiment setup
To evaluate the performance of the proposed method, simulation and ex vivo experiments were conducted. In the simulation, PA signals were generated by using a k-wave toolbox ; an absorber with the size of 0.09 mm was placed at the depth of 20 mm and scanned by a 5-MHz ultrasound linear array with 128 elements. The element pitch was 0.3 mm. The generated PA signals propagated through a medium with a sound speed of 1500 m/s and received by the array of which fraction bandwidth was 80%.
In addition to the simulation, the ex vivo experiments were conducted. The system configuration for ex vivo experiments is shown in Fig. 2 . The microcalcification-contained breast specimen was fixed on PVC phantom with rubber band and immersed in 0.9% saline solution. RF channel data were acquired with a commercial ultrasound scanner equipped with a research package by using a 7-MHz linear array transducer (SonixTouch, SonixDAQ, and L14-5/38, Ultrasonix Corp., Canada). The acquisition system was synchronized with Q-switch trigger output of an OPO laser system pumped by Nd:YAG laser (Surelite OPO Plus and Surelite III-10, Continuum Inc., USA) using a function generator (AFG3252, Tektronix Corp., USA) with frame rate of 10 Hz. The transducer was positioned at the top away 30 mm from the specimen, and laser energy was delivered at the side. A 7-ns laser pulse used to generate the PA signal had the pulse energy of 19 mJ/cm2; since the optimum wavelength for breast microcalcifications was found to be 690 nm in the previous study , a laser pulse with the wavelength was utilized. In the experiments, RF channel data of both PA and US were acquired, and image reconstruction was performed based on DAS-BF for US imaging. The PA imaging of both simulation and ex vivo experiments was conducted by the adaptive beamforming based on the CF to reduce the artifacts (e.g., degradation of spatial resolution) stemming from one-way focusing [7, 8]. The iterative processing was conducted varying the sound speed from 1460 m/s to 1560 m/s at a 1 m/s increment.
3. Results and discussion
Figure 3(a) shows a PA image obtained by the simulation; ROI for estimating sound speed was indicated. As shown in Fig. 3(b), the maximum of the CF in the ROI was achieved at the sound speed of 1500 m/s. This sound speed is equal to the true value used in the simulation. The −6 dB lateral resolution of the absorber image was measured as a function of sound speed and the minimum beam width of 0.3 mm was obtained at the true sound speed of 1500 m/s as shown in Fig. 3(c). The −6 dB lateral beam width was gradually broadened as the difference between sound speed and the true one was increased. The simulation results indicate that the proposed method is capable of estimating the true sound speed and improving lateral resolution of PA images.
Figure 4 presents the ex vivo experimental results; normalized maximal CF values corresponding to each sound speed are plotted in Fig. 4(a) and the ROI for estimating optimal sound speed is indicated by a white box in Fig. 4(b). As shown in Fig. 4(a), the maximal CF was found at a sound speed of 1496 m/s, which is similar to the simulation result. Although the medium contained fat (or soft tissue) in the specimen, the main composition was saline. Therefore, the optimal sound speed came close to the sound speed in saline solution, which can be computed by Eq. (4), the sound speed in 0.9% saline solution was calculated to be 1503.8 m/s when T = 24 þC. This was similar to the estimated sound speed, i.e., 1496 m/s. Figures 4(b) and (c) show the reconstructed PA images registered on US images with the commonly used sound speed (i.e.,1540 m/s) and the estimated one (i.e., 1496 m/s), respectively. The dynamic ranges were 30 dB and 60 dB for the PA and the US images, respectively. Under visual assessment, it is easily recognized that the sound speed of 1496 m/s (Fig. 4(b)) provides higher distinction of individual microcalcifications than the one of 1540 m/s.
The lateral beam profiles of the microcalcification (indicated by an arrow in Fig. 4(b)) were measured from the PA images and presented in Fig. 5(a) . The image formed with the sound speed of 1496 m/s provided narrower lateral beam width than that with 1540 m/s; the −6 dB lateral beam width was 0.62 mm in the case of the sound speed of 1496 m/s, while it was 1.40 mm in the case of 1540 m/s. These results can be explained with the fact that the image reconstruction with the optimal sound speed allows for a stark difference between CFs for the main lobe and sidelobes as shown in Fig. 5(b), compared to that with the wrong sound speed.
When the −6 dB lateral beam widths from the same microcalcification were measured as a function of sound speeds (see Fig. 6 ), it was learned that the lateral resolution was deteriorated more and more as the sound speed went away from the optimal one. In addition, the signal-to-noise ratio (SNR), defined as a ratio of the maximum intensity from 1496 m/s to that from 1540 m/s, was improved by 3.7 dB due to higher efficacy in beamforming and CF weighting.
Phase distortion (i.e., phase aberration) due to the variation of sound speed is a function of imaging depth. Therefore, the accurate compensation for the phase distortion may be conducted by measuring phase aberration or sound speed variation along imaging depth. This can be done by computing cross-correlation between RF data at each channel, which requires considerable computational cost . Although the proposed method is not for measuring accurate sound speed along imaging depth, it can provide an optimal sound speed with which the best spatial resolution is achieved in the ROI. The proposed method may be more useful in combined PA/US imaging because the ROI for PA imaging is usually confined over a large area of US imaging.
In the calculation of CF for ultrasound imaging, there are two possible error sources that under- or over-estimate the value: low input SNR and strong reflectors positioned outside ROI . The low input SNR is generally problematic in ultrasound imaging when CF is calculated for point scatterers and anechoic cysts of which amplitude is either similar to or a few times higher than noise level. Also, the sidelobes from strong reflectors outside ROI may affect the accuracy of CF when the ROI contains only weak reflectors. However, the proposed method is robust against those error sources because the sound speed estimation is conducted using PA signals from strong absorbers; in particular, this is true in the PA imaging for microcalcifications. The amplitude of PA signals is generally 30 dB higher than noise level, so that the low input SNR is no longer any issue. Since strong absorbers are the main concern in PA imaging, furthermore, ROI should contain the strong absorbers.
In a general ultrasound scanner, beamforming algorithms are implemented in hardware such as an application-specific integrated circuit (ASIC) and a field programmable gate array (FPGA). Therefore, each beamformed sample can be obtained at every operating clock period after a certain amount of latency. If PA signals are sampled at 60 MHz, beamformation is performed at 120 MHz, and the size of ROI is 30 mm by 30 mm, the reconstruction of one PA image with the calculation of CF can be conducted in 1 ms. Note that the width of ROI is determined by the number of scalines (100 in this example) and the array element pitch (0.3 mm). The frame rate of the proposed method is limited by the pulse repetition frequency of a Nd:YAG laser system of which maximum is currently 15 Hz, i.e., 66.7 ms. In this example, therefore, the possible number of iterations is of 66 to estimate the optimal sound speed. If the increment of sound speed in the iteration processing is less than 1.5 m/s, the proposed method can be implemented in real time. Although we used the 1 m/s increment of sound speed, the number of iterations can be reduced with coarser sound speed steps, e.g., 10 m/s without degrading lateral resolution. In addition, the standard optimization methods such as the golden search algorithm can be utilized to further reduce the number of iterations and the processing time .
In this paper, we proposed the method for an optimal sound speed estimation, which was based on maximizing CF in ROI for PA imaging. The proposed method was verified through the simulation and the ex vivo experiments with the microcalcification-contained breast specimen. The results demonstrated that proposed method can enhance PA image quality, and is useful in observation of breast microcalcifications.
This work was supported by International Collaborative R&D Program (2010-TD-500409-001) funded by the Ministry of Knowledge Economy (MKE), South Korea.
References and links
1. H. F. Zhang, K. Maslov, G. Stoica, and L. V. Wang, “Functional photoacoustic microscopy for high-resolution and noninvasive in vivo imaging,” Nat. Biotechnol. 24(7), 848–851 (2006). [CrossRef] [PubMed]
2. R. I. Siphanto, K. K. Thumma, R. G. M. Kolkman, T. G. van Leeuwen, F. F. M. de Mul, J. W. van Neck, L. N. A. van Adrichem, and W. Steenbergen, “Serial noninvasive photoacoustic imaging of neovascularization in tumor angiogenesis,” Opt. Express 13(1), 89–95 (2005). [CrossRef] [PubMed]
3. M. Xu and L. V. Wang, “Photoacoustic imaging in biomedicine,” Rev. Sci. Instrum. 77(4), 0411011 (2006). [CrossRef]
4. T. N. Erpelding, C. Kim, M. Pramanik, L. Jankovic, K. Maslov, Z. Guo, J. A. Margenthaler, M. D. Pashley, and L. V. Wang, “Sentinel lymph nodes in the rat: noninvasive photoacoustic and US imaging with a clinical US system,” Radiology 256(1), 102–110 (2010). [CrossRef] [PubMed]
5. X. Wang, J. B. Fowlkes, J. M. Cannata, C. Hu, and P. L. Carson, “Photoacoustic imaging with a commercial ultrasound system and a custom probe,” Ultrasound Med. Biol. 37(3), 484–492 (2011). [CrossRef] [PubMed]
6. T. Harrison and R. J. Zemp, “The applicability of ultrasound dynamic receive beamformers to photoacoustic imaging,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control 58(10), 2259–2263 (2011). [CrossRef] [PubMed]
9. C. Yoon, Y. Lee, J. H. Chang, T.-K. Song, and Y. Yoo, “In vitro estimation of mean sound speed based on minimum average phase variance in medical ultrasound imaging,” Ultrasonics 51(7), 795–802 (2011). [CrossRef] [PubMed]
10. B. E. Treeby, T. K. Varslot, E. Z. Zhang, J. G. Laufer, and P. C. Beard, “Automatic sound speed selection in photoacoustic image reconstruction using an autofocus approach,” J. Biomed. Opt. 16(9), 090501 (2011). [CrossRef] [PubMed]
12. K. P. Köstli and P. C. Beard, “Two-dimensional photoacoustic imaging by use of Fourier-transform image reconstruction and a detector with an anisotropic response,” Appl. Opt. 42(10), 1899–1908 (2003). [CrossRef] [PubMed]
14. J. Kang, E. K. Kim, J. Y. Kwak, Y. Yoo, T.-K. Song, and J. H. Chang, “Optimal laser wavelength for photoacoustic imaging of breast microcalcification,” Appl. Phys. Lett. 99(15), 153702 (2011). [CrossRef]
15. C. R. Hill, J. C. Bamber, and G. Haar, “Physical principles of medical ultrasonics (John Wiley and Sons, 2004), Chap. 5.
16. S. W. Flax and M. O’Donnell, “Phase-aberration correction using signals from point reflector and diffuse scatterers: basic principles,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control 35(6), 758–767 (1988). [CrossRef]