Photoacoustic spectrum analysis at either 1200 nm or 532 nm can allow differentiation of fatty from normal liver tissue in a mouse model by showing the microstructural changes associated with lipid and hemoglobin in the liver tissue.
Abstract
Purpose
To investigate the use of photoacoustic (PA) spectrum analysis (PASA) to identify microstructural changes corresponding to fat accumulation in mouse livers ex vivo and in situ.
Materials and Methods
The laboratory animal protocol for this work was approved by the university committee on use and care of animals. Six mice with normal livers and six mice with fatty livers were examined ex vivo with a PA system at 1200 nm, and nine similar pairs of mice were examined at 532 nm. To explore the feasibility of this technique for future study in an in vivo mouse model, an additional pair of normal and fatty mouse livers was scanned in situ with an ultrasonographic (US) and PA dual-modality imaging system. The PA signals acquired were analyzed by using the proposed PASA method. Results of the groups were compared by using the Student t test.
Results
Prominent differences between the PASA parameters from the fatty and normal mouse livers were observed. The analysis of the PASA parameters from six normal and six fatty mouse livers indicates that there are differences of up to 5 standard deviations between the PASA parameters of the normal livers and those of the fatty livers at 1200 nm; for parameters from nine normal and nine fatty mouse livers at 532 nm, the differences were approximately 2 standard deviations (P < .05) for each PASA parameter.
Conclusion
The results supported our hypothesis that the PASA allows quantitative identification of the microstructural changes that differentiate normal from fatty livers. Compared with that at 532 nm, PASA at 1200 nm is more reliable for fatty liver diagnosis.
© RSNA, 2014
Introduction
Biopsy is regarded as the reference standard for diagnosis of many diseases (1–3) because the method directly reveals the histologic changes in biologic tissue. However, the invasive nature and the length of time it takes to perform a biopsy make it less desirable for many conditions (4). The diagnosis and treatment monitoring of many diseases can be drastically improved by using a noninvasive imaging modality that facilitates the quantification of the histologic microstructures with adequate sensitivity and specificity. In photoacoustic (PA) imaging, light from a pulsed laser is used to illuminate a biologic sample. The light energy deposition in the tissue leads to an instant temperature rise and thermoelastic expansion, which induce ultrasonographic (US) waves (ie, PA signals). Although they are very weak in intensity, the PA signals can be collected by using US transducers or other US detectors for later reconstruction of an image of the sample. Compared with conventional optical imaging, the spatial resolution of PA imaging is not limited by the strong light diffusion, but instead, is determined by the scanning geometry and the receiving bandwidth in detecting PA signals. Therefore, PA imaging has a major advantage over existing optical modalities and can render detailed features in optically scattering tissue even when the imaging depth is beyond the optical mean free path. As an example, PA imaging of the human breast has been achieved recently with satisfactory spatial resolution at a depth of up to 5 cm from the skin surface (5).
Almost all previous studies in PA imaging have been focused on the intensity of the PA signal from biologic tissue as an indication of macroscopic optical absorbance (6). Our group demonstrated previously (7–10) that the frequency domain power distribution (power spectrum) of the broadband radiofrequency PA signals also encodes the textural information in the regions of interest. Authors of earlier studies (11,12) have investigated extensively signal power spectrum analysis in US imaging. US spectrum analysis has been used to study the intensity attenuation and frequency or phase shift of the backscattered US waves and periodicity captured by the spectrum and allows discrimination of microscopic features in biologic tissue (13–17). Similar to US spectrum analysis, PA spectrum analysis (PASA) could allow evaluation of the intensity and, more importantly, the “pitch” or frequency of the PA signals.
Normal mouse livers possess compact, homogeneous cell structures and abundant red blood cells in the intercellular sinusoids but no noticeable accumulation of fat, as shown in Figure E1 (online). When fat accumulates in experimental models of obesity, the number of red blood cells per unit area decreases. As shown in Figure E2 (online), the molecular vibrational absorption peak of the carbon-hydrogen bond in a lipid molecule at approximately 1200 nm (18) and the hemoglobin absorption spectrum peak at approximately 532 nm (19) make these two wavelengths promising for studying the lipid and blood distribution in liver tissue. Several hypotheses can be made on the basis of observations of histologic images in Figure E1 (online): (a) Higher spectral intercept and midband-fit values in a first-order model will be observed at 532 nm in normal liver tissue because of the higher blood content in normal livers compared with fatty livers. (b) Higher spectral intercept and midband-fit values will be observed at 1200 nm in fatty liver tissue because of the increased lipid content in fatty livers compared with that in normal livers. (c) Negative slopes with smaller absolute values will be observed in fatty livers at both 1200 nm and 532 nm because the heterogeneous tissue structure in fatty livers generates high-frequency components, whereas the more homogeneous normal livers primarily produce low-frequency PA signals at both 1200 nm and 532 nm.
Our purpose was to investigate the capability of PASA to allow identification of the microstructure changes corresponding to fat accumulation in mouse livers through ex vivo and in situ experiments.
Materials and Methods
The laboratory animal protocol for this work was approved by the university committee on use and care of animals of the University of Michigan.
Animal Model Preparation
C57BL/6 J wild-type mice (Jackson Laboratory, Bar Harbor, Maine) were used in this study. The obese group was fed with a chow diet for the first 8 weeks, followed by 60% fat diet (research diet, D12492) for the 12 weeks thereafter. The control group was fed with the chow diet for 20 weeks. Both groups were sacrificed at the end of the 20th week. The obese and the control groups each included 16 mice, with six and nine mice for the ex vivo experiments at 1200 nm and 532 nm, respectively, and one mouse for the in situ experiment.
Experimental Setups
Comparable to that of radiofrequency US signals, the power spectrum of the radiofreqency PA signals in decibels can be approximated by a first order model. The detailed description of such a model can be found in our previous work (8) (for one-dimensional imaging) and Appendix E1 of this work (online; for two-dimensional [2D] imaging). Three parameters, including intercept, midband fit and slope of the first order model, will afterward be extracted (7,8,11). Intercept is the magnitude of the linear model at zero frequency, representing the low-frequency components of the signal power spectrum. Midband fit is the magnitude of the linear model at the center frequency of the fitting range, representing the averaged signal spectrum magnitude in the entire fitting range. Intercept and midband fit both reflect the macroscopic absorption of the biologic tissue. Slope represents the distribution of the frequency components of the radiofrequency PA signals. Higher slope value indicates more high-frequency components and equivalently more heterogeneous tissue texture, whereas lower slope value indicates more homogeneous tissue features. In this study, to achieve desirable signal-to-noise ratio, PA signals were averaged 100 times in both the ex vivo and the in situ experiments.
Ex vivo experiment.—The home-built setup for the ex vivo PA imaging experiment is illustrated in Figure 1a.The optical illumination was generated by a tunable Optical Parametric Oscillator laser (Vibrant B, Opotek, Carlsbad, Calif) pumped by the second harmonic output of an Nd:YAG pulsed laser (Brilliant B, Quantel, Bozeman, Mt). For imaging at 1200 nm, the laser beam with 18 mJ per pulse and pulse duration of 8 nsec was collimated to 13 mm in diameter to illuminate the whole area of a sample piece, achieving an averaged light fluence of 13.6 mJ/cm2 on the sample surface (within the safety limit of the American National Standards Institute regulations). The sample piece was cut from an intact lobe of a mouse liver into a disc shape with a diameter of 13 mm and a thickness of 1–2 mm. For the experiment at 532 nm, we used the light beam from the pumping laser at the energy level of 200 mJ per pulse. After the beam was expanded to 50 mm in diameter, the average light fluence on the sample surface was 10.2 mJ/cm2. A needle hydrophone (HNC-1500, ONDA, Sunnyvale, Calif), with a detection bandwidth of 20 MHz centered at 10 MHz with a frequency-dependent response variation of ± 3 dB was used to acquire PA signals. To perform 2D tomographic imaging of an ex vivo liver, a circular scan of the needle hydrophone around the tissue with a scanning radius of 17.5 mm was conducted as described in the literature (20,21). For ultrasonic coupling, both the hydrophone and the liver tissue were placed in a tank of water. The liver tissue was fixed in a cylindrical sample holder made from porcine gel. After amplification by a total of 60 dB by using a preamplifier (30 dB, AH-2010, ONDA) and a low-noise amplifier (30 dB 5072PR, Parametrics, Waltham, Mass), the PA signal received by the hydrophone was recorded with a digital oscilloscope (TDS 540, Tektronix, Beaverton, Ore) at a sampling rate of 250 MHz. A personal computer with LabView control synchronized the laser firing, the data acquisition of the oscilloscope, and the rotation of the hydrophone. PA signals were acquired at 240 angular positions evenly distributed around the sample.
Figure 1a:
Illustration shows experiment setups for (a) ex vivo and (b) in situ imaging of mouse liver. GPIB = general purpose interface bus.
After down-sampling to 50 MHz, the PA signals were used to reconstruct an image of the sample with a modified back-projection algorithm (20,21). Appendix E1 (online) provides details about the methods for power spectrum calculation in two dimensions. For the data acquired at 1200 nm, we only looked at the spectral range of 0.3–4.6 MHz, corresponding to approximately the 20-dB level in the data bandwidth. The relatively narrow spectral range was selected to validate that PASA can allow evaluation of the microstructures in the tissue by using only part of the signal power spectrum (9). The frequency range started at 0.3 MHz because of the discretization of the frequency axes. Starting from 0.3 MHz instead of 0 MHz also partially excluded the low-frequency signal components produced by the light illumination on the background porcine gel. Before we processed the data acquiredat 532 nm, we discarded the lowest 2- MHz spectral range to avoid the strong, low-frequency PA signals generated from the light illumination on the background porcine gel. Then the power spectrum was analyzed in the wider range of 2–8.3 MHz including the 30-dB level.
In situ experiment.—In the in situ experiment, a customized US and PA dual-modality imaging system was used for rapid data acquisition, as shown in Figure 1b. This system was based on a US platform (Verasonics, Redmond, Wash) and a 128-element linear array (L7–4, Philips Healthcare, Andover, Mass) working at a sampling frequency of 20 MHz. Powered by a graphics processing unit card that facilitates parallel computation, this system can perform PA and US imaging of the same object simultaneously at a frame rate of 10 Hz, which is limited by the pulse repetition rate of the laser (22). Because our laser provides limited output energy of 18 mJ per pulse at 1200 nm, we opened the abdominal cavity of the scanned mouse by performing bilateral incisions to facilitate the exposure of the liver to avoid light attenuation in the belly, which reduces the signal-to-noise ratio. The laser beam with a diameter of 13 mm illuminated the liver surface with an average light fluence of 13.6 mJ/cm2 for the scans at both 1200 nm and 532 nm. The alignment of the imaging plane to the liver tissue was first confirmed by means of US imaging. Q-switch of the laser triggered the US platform for PA signal acquisition. The signals collected by the US system were stored in the controlling computer for image reconstruction, display, and later, PASA. The pitch of 298 μm of the L7–4 probe was equivalent to the sampling frequency of approximately 0.19 MHz, assuming a 1560 μm/μsec speed of sound (23). Because of the low sampling rate along the lateral direction compared with that along the axial direction, the in situ experimental data were processed with the one-dimensional PASA method described in our previous work (7). Each A-line of the beam formed radiofrequency signals in the regions of interest (approximately 4 mm × 7 mm; Fig 2) was analyzed by using the Pwelch method with a one-dimensional sliding window of 1.6 μsec. The regions of interest were confirmed by a physician (Z.M., with 9 years of experience in metabolism in cells, tissues, and organisms). The power spectra were afterward calibrated by the frequency response of the US probe according to the method described by Kumon et al (7). The calibrated spectra were quantified in a range of 1.5–7.5 MHz including the 15-dB bandwidth of the probe. One normal and one obese mouse were examined at both 1200 nm and 532 nm. Each PASA parameter was averaged in the regions of interest.
Figure 1b:
Illustration shows experiment setups for (a) ex vivo and (b) in situ imaging of mouse liver. GPIB = general purpose interface bus.
Figure 2:
PASA images show the spatial distribution of the pixel-wise PASA parameters from in situ experiments at 1200 nm and 532 nm. Parameter distributions were calculated at each step of sliding window with Pwelch algorithm. Similar to those in ex vivo results, PASA parameters at 1200 nm, including intercept, midband fit, and slope of normal tissue have lower values than those of fatty tissue. At 532 nm, normal liver shows higher intercept and midband-fit values but lower slope values.
Statistical Analysis
The intercept, midband-fit, and slope values acquired from the ex vivo experiment at 1200 nm and 532 nm were examined by using a two-tailed student t test. A P value of .05 was considered to indicate a significant difference. The statistical analysis was conducted with the built-in statistical functions in Matlab R2011b (MathWorks, Natick, Mass).
Results
Ex Vivo Experiment
The normal and fatty tissue types could be reliably identified at either wavelength with any of the three PASA parameters (Figs 3, 4). There were differences of up to 5 standard deviations (ie, the slope) between the PASA parameters from the fatty and the normal livers when they were imaged at 1200 nm. These differences were approximately 2 standard deviations (ie, the midband fit and the slope) when the livers were imaged at 532 nm.
Figure 3:
PASA images show spatial distribution of the pixelwise PASA parameters of typical sample specimens at 1200 and 532 nm. Parameter distributions were calculated at each step of sliding window with 2D Pwelch algorithm. At 1200 nm, normal liver shows lower intercept, midband-fit, and slope values compared with fatty liver. At 532 nm, normal liver shows higher intercept and midband-fit values but lower slope values.
Figure 4:
Graphs show statistical analysis of the PASA parameters from the ex vivo experiment at 1200 nm and 532 nm, respectively. o and x = data point from normal and fatty livers, respectively, and 
and 
= averaged PASA parameter of normal and fatty livers, respectively. P values were calculated by using two-tailed Student t test.
In Situ Experiment
The comparison between the PASA parameters from the normal and the fatty livers in situ in Figure 2 and the Table validated our hypothesis that at 1200 nm, the fatty liver possessed higher spectral parameters including slope, intercept, and midband fit than did normal livers. At 532 nm, the fatty liver had high slope values yet lower intercept and midband-fit values compared with those of the normal liver.
Averaged PASA Parameters in the Regions of Interest in the in Situ Model
Discussion
US and PA spectrum analyses have similar procedures, yet the two methods are somewhat different. US spectrum analysis characterizes biologic tissue with narrow-band US waves backscattered due to the acoustic impedance mismatch, whereas PASA quantifies tissue microstructures by accessing the optical absorption contrast and by analyzing the broadband PA signals originating in the regions of interest. The reconstructed PA images based on the back-projection algorithm are fundamentally the 2D beam-formed wideband radiofrequency PA signals. The 2D spectrum analysis approach was introduced to evaluate the microstructures in the 2D scanning plane. In the 2D PASA, the sliding window for calculating the 2D power spectrum of each subsection of the PA image must be square, and the sampling rates in the orthogonal dimensions must be comparable to facilitate the slope calculation. Otherwise, the PASA must be reduced to one dimension along the dimension with the higher sampling rate to be similar to that for the in situ experiment. The square sliding window in 2D PASA also conforms to the presumption that the tissue texture is isotropic and ensures that the tissue texture is evaluated equally in both dimensions. In our study, the quantification of the optical absorption of the tissue represented by the intercept and the midband fit is still relative due to the lack of knowledge on the accurate light energy deposition. The quantitative comparability of these two PASA parameters relies on the uniform light fluence in the paired imaging experiment on fatty and normal livers. In comparison with the intercept and the midband-fit, the slope is least dependent on the light fluence and, therefore, potentially could be used in the future to quantify the microstructures in the tissue.
The PASA that was focused on the total hemoglobin content in the liver was performed at a wavelength of 532 nm, benefiting from the very stable energy output and the high beam quality of the pumping laser working at 532 nm. The increased overlapping of the data for the normal and fatty livers at 532 nm compared with that at 1200 nm may have been because blood cells with an optical absorption peak at 532 nm exist in the sinusoids of both normal and fatty livers, whereas the lipid droplets exist only in fatty livers. In the future, especially for noninvasive imaging of the liver in vivo, the laser light in the near-infrared spectral region between 700 nm and 900 nm could be a better option because better imaging depth can be achieved in this optical window. Moreover, when multiple laser wavelengths corresponding to the optical absorption peaks of oxygenated and deoxygenated hemoglobin can be used, PASA may be helpful in evaluating the spatial distributions of these two major forms of hemoglobin in biologic tissues.
All the fatty livers were from extremely obese mice. Our future investigations will include the sensitivity of PASA with respect to the progression of steatosis. We also expect that, at earlier stages, multivariate analysis including all three or any two of the spectral parameters could work better when a single PASA parameter does not suffice for the identification of fatty livers. Future investigations will also be extended to more advanced stages of fatty liver disease such as fibrosis, which could be characterized by using PASA at the optical fingerprint of collagen around 1350 nm because fibrosis is associated with elevated collagen content (24).
The ultimate limit of PASA could be the signal bandwidth of the scanning system and the penetration depth of the laser energy. Studies on human breast, which is rich in adipose tissue, demonstrated that the PA imaging depth reach 5 cm when the laser light is in the near-infrared spectrum region (25). This imaging depth is sufficient for the study of small animals including rats and mice and may allow for scanning of patients when the liver is imaged from the side to avoid the thick subcutaneous fat at the front of the abdomen (24). In the in situ experiment, because our laser was not sufficiently powerful, we tried to avoid any attenuation of light before it reached the liver by opening the abdominal cavity. When a more powerful laser is available in the future, no surgery would be necessary for small animals and the whole procedure could be conducted in a noninvasive manner. However, utility in obese patients must be proved.
Practical applications: In our study, we investigated the feasibility of differentiating fatty and normal livers by using PASA. The ex vivo and the in situ experiments in a mouse model at both 1200-nm and 532-nm wavelengths validated our hypotheses on the relationship between the spectral parameters and the microstructures in mouse livers, although 1200-nm illumination appears to be more promising for fatty liver identification. PASA holds promise as an in vivo and noninvasive method to identify the microstructures in liver tissue to aid liver disease diagnosis.
Advances in Knowledge
■ Photoacoustic spectrum analysis (PASA) of mouse liver tissue at either 1200 nm or 532 nm can allow quantification of the microscopic distribution of lipid and hemoglobin.
■ A study of six normal and six fatty mouse livers at 1200 nm showed differences of up to 5 standard deviations (ie, the slope) between the PASA parameters from the fatty and normal livers.
■ A study of nine normal and nine fatty mouse livers at 532 nm showed differences of up to 2 standard deviations (ie, the midband fit and the slope) between the PASA parameters from the fatty and the normal livers.
APPENDIX
SUPPLEMENTAL FIGURES
Acknowledgments
Acknowledgments
Thanks to Jinxing Cheng, PhD, from Purdue University for providing the PA spectrum data. We would also like to thank J. Brian Fowlkes, PhD, from the University of Michigan for his support to this work.
Received April, 1 2013; revision requested June 4; revision received September 3; accepted October 1; final version accepted October 22.
Supported by the National Natural Science Foundation of China (grants 11028408 and 61201425) and Samsung GRO Program.
Funding: This research was supported by the National Institutes of Health (grant R01AR060350).
Disclosures of Conflicts of Interest: G.X. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: author plans to file a patent for this technique. Other relationships: none to disclose. Z.X.M. No relevant conflicts of interest to disclose. J.D.L. No relevant conflicts of interest to disclose. J.Y. No relevant conflicts of interest to disclose. P.L.C. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: none to disclose. Other relationships: author has a patent pending. B.J. No relevant conflicts of interest to disclose. X.W. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: none to disclose. Other relationships: author plans to file a patent for this technique.
Abbreviations:
- PA
- photoacoustic
- PASA
- PA spectrum analysis
- 2D
- two dimensional
References
- 1.Rabbani F, Stroumbakis N, Kava BR, Cookson MS, Fair WR. Incidence and clinical significance of false-negative sextant prostate biopsies. J Urol 1998;159(4):1247–1250. [PubMed] [Google Scholar]
- 2.Gortzak-Uzan L, Jimenez W, Nofech-Mozes S, et al. Sentinel lymph node biopsy vs. pelvic lymphadenectomy in early stage cervical cancer: is it time to change the gold standard? Gynecol Oncol 2010;116(1):28–32. [DOI] [PubMed] [Google Scholar]
- 3.Joy D, Thava VR, Scott BB. Diagnosis of fatty liver disease: is biopsy necessary? Eur J Gastroenterol Hepatol 2003;15(5):539–543. [DOI] [PubMed] [Google Scholar]
- 4.Cadranel J-F, Rufat P, Degos F. Practices of liver biopsy in France: results of a prospective nationwide survey. For the Group of Epidemiology of the French Association for the Study of the Liver (AFEF). Hepatology 2000;32(3):477–481. [DOI] [PubMed] [Google Scholar]
- 5.Xie Z, Hooi FM, Fowlkes JB, Pinsky RW, Wang X, Carson PL. Combined photoacoustic and acoustic imaging of human breast specimens in the mammographic geometry. Ultrasound Med Biol 2013;39(11):2176–2184. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Wang LV, Hu S. Photoacoustic tomography: in vivo imaging from organelles to organs. Science 2012;335(6075):1458–1462. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Kumon RE, Deng CX, Wang X. Frequency-domain analysis of photoacoustic imaging data from prostate adenocarcinoma tumors in a murine model. Ultrasound Med Biol 2011;37(5):834–839. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Xu G, Dar IA, Tao C, Liu X, Deng CX, Wang X. Photoacoustic spectrum analysis for microstructure characterization in biological tissue: A feasibility study. Appl Phys Lett 2012;101(22):221102–221105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Yang Y, Wang S, Tao C, Wang X, Liu X. Photoacoustic tomography of tissue subwavelength microstructure with a narrowband and low frequency system. Appl Phys Lett 2012;101(3):034105. [Google Scholar]
- 10.Wang S, Tao C, Wang X, Liu X. Quantitative detection of stochastic microstructure in turbid media by photoacoustic spectral matching. Appl Phys Lett 2013;102(11):114102. [Google Scholar]
- 11.Lizzi FL, Greenebaum M, Feleppa EJ, Elbaum M, Coleman DJ. Theoretical framework for spectrum analysis in ultrasonic tissue characterization. J Acoust Soc Am 1983;73(4):1366–1373. [DOI] [PubMed] [Google Scholar]
- 12.Lizzi FL, Astor M, Feleppa EJ, Shao M, Kalisz A. Statistical framework for ultrasonic spectral parameter imaging. Ultrasound Med Biol 1997;23(9):1371–1382. [DOI] [PubMed] [Google Scholar]
- 13.Meyer CR, Herron DS, Carson PL, et al. Estimation of ultrasonic attenuation and mean backscatterer size via digital signal processing. Ultrason Imaging 1984;6(1):13–23. [DOI] [PubMed] [Google Scholar]
- 14.Ossant F, Lebertre M, Pourcelot L, Patat F. Ultrasonic characterization of maturation of fetal lung microstructure: an animal study. Ultrasound Med Biol 2001;27(2):157–169. [DOI] [PubMed] [Google Scholar]
- 15.Mia RS, Loew MH, Wear KA, Wagner RF. Mean scatterer spacing estimation using the complex cepstrum. In: Hanson KM, ed. Proceedings of SPIE: medical imaging 1997—image processing. Vol 3034. Bellingham, Wash: SPIE–The International Society for Optical Engineering, 1997; 713–722. [Google Scholar]
- 16.Huang K, Ta D, Wang W, Le LH. Simplified inverse filter tracking algorithm for estimating the mean trabecular bone spacing. IEEE Trans Ultrason Ferroelectr Freq Control 2008;55(7):1453–1464. [DOI] [PubMed] [Google Scholar]
- 17.Lizzi FL, Astor M, Liu T, Deng C, Coleman DJ, Silverman RH. Ultrasonic spectrum analysis for tissue assays and therapy evaluation. Int J Imaging Syst Technol 1997;8(1):3–10. [Google Scholar]
- 18.Wang HW, Chai N, Wang P, et al. Label-free bond-selective imaging by listening to vibrationally excited molecules. Phys Rev Lett 2011;106(23):238106. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Prahl S. Optical Absorption of Hemoglobin. http://omlc.ogi.edu/spectra/hemoglobin/index.html. Published December 15, 1999. Accessed March 20, 2013.
- 20.Wang X, Pang Y, Ku G, Xie X, Stoica G, Wang LV. Noninvasive laser-induced photoacoustic tomography for structural and functional in vivo imaging of the brain. Nat Biotechnol 2003;21(7):803–806. [DOI] [PubMed] [Google Scholar]
- 21.Wang X, Pang Y, Ku G, Stoica G, Wang LV. Three-dimensional laser-induced photoacoustic tomography of mouse brain with the skin and skull intact. Opt Lett 2003;28(19):1739–1741. [DOI] [PubMed] [Google Scholar]
- 22.Yuan J, Xu G, Yu Y, et al. Real-time photoacoustic and ultrasound dual-modality imaging system facilitated with graphics processing unit and code parallel optimization. J Biomed Opt 2013;18(8):86001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Madjar H, Mendelson E. The practice of breast ultrasound: techniques, findings, differential diagnosis. 2nd ed. New York, NY: Thieme, 2008; 3. [Google Scholar]
- 24.Palmeri ML, Wang MH, Rouze NC, et al. Noninvasive evaluation of hepatic fibrosis using acoustic radiation force-based shear stiffness in patients with nonalcoholic fatty liver disease. J Hepatol 2011;55(3):666–672. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Xie Z, Lee WM, Hooi FM, et al. Combined photoacoustic and ultrasound imaging of human breast in vivo in the mammographic geometry. In: Oraevsky AA, Wang LV, eds. Proceedings of SPIE: medical imaging 2013—photons plus ultrasound: imaging and sensing 2013. Vol 8581. Bellingham, Wash: SPIE–The International Society for Optical Engineering, 2013; 85813D. [Google Scholar]
- 26.Welch P. The use of fast Fourier transform for the estimation of power spectra: A method based on time averaging over short, modified periodograms. IEEE Trans Audio Electroacoust 1967;15(2):70–73. [Google Scholar]
- 27.van Veen RL, Sterenborg HJ, Pifferi A, Torricelli A, Cubeddu R. Determination of VIS-NIR absorption coefficients of mammalian fat, with time- and spatially resolved diffuse reflectance and transmission spectroscopy. Washington, DC: Optical Society of America, 2004; SF4. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.