J. Biophotonics 8, No. 5, 408–414 (2015) / DOI 10.1002/jbio.201400047

FULL ARTICLE

Brillouin spectroscopy as a new method of screening for increased CSF total protein during bacterial meningitis Zachary Steelman, Zhaokai Meng, Andrew J. Traverso, and Vladislav V. Yakovlev Biomedical Engineering Department, Texas A&M University, College Station, TX 77843, USA Received 16 April 2014, revised 9 May 2014, accepted 16 May 2014 Published online 15 July 2014

Key words: spectrum analysis, meningitis, bacterial, pathology, molecular, optics and photonics Bacterial meningitis is a disease of pronounced clinical significance, especially in the developing world. Immediate treatment with antibiotics is essential, and no single test can provide a conclusive diagnosis. It is well established that elevated total protein in cerebrospinal fluid (CSF) is associated with bacterial meningitis. Brillouin spectroscopy is a widely used optical technique for noninvasive determination of the elastic moduli of materials. We found that elevated protein levels in CSF alter the fluid elasticity sufficiently to be measurable by Brillouin spectroscopy, with model healthy and diseased fluids distinguishable to marked significance (P = 0.014), which increases with sample concentration by dialysis.

Typical raw output of a 2-stage VIPA Brillouin spectrometer: inelastically scattered Brillouin peaks (arrows) and elastically scattered incident radiation (center cross).

1. Introduction Bacterial meningitis is a challenging diagnosis for most physicians due to the non-specificity of symptoms, particularly in small children [1]. Even with early diagnosis and treatment, a 10% mortality rate persists, often within a day or two of symptom onset [2, 3]. If no treatment is available, as is the case in many parts of the developing world (where meningitis is up to ten times more prevalent), the mortality * Corresponding author: e-mail: [email protected] © 2014 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

rate increases to as high as 80% [3]. Furthermore, bacterial meningitis causes significant complications for 50% of survivors [4]. A lumbar puncture is generally performed immediately upon suspicion of meningitis in order to obtain a sample of spinal fluid [5]. Rapid diagnosis and immediate treatment with antibiotics greatly reduces the mortality rate of bacterial meningitis [6], however most existing diagnostic tests of the spinal fluid are either slow (such as bacterial cultures, which require at least 24 hours [7]) or destruc-

J. Biophotonics 8, No. 5 (2015)

tive to at least a portion of the CSF sample (generally, through mixing with laboratory indicators or heavy dilution [8, 9]). Because many of these traditional methods require alteration of the CSF fluid, volume of sample is restricted for other available tests, significant time is lost by the physician or lab technician, and the cost of chemical dyes and indicators in incurred. However, in virtually all cases of bacterial meningitis, protein concentrations in the CSF are elevated [10]. Therefore, detecting an increase in the total protein content of spinal fluid, along with identification of common symptoms, would be indicative of bacterial meningitis, allowing physicians to begin an immediate treatment regimen of antibiotics. Many common methods of detecting total protein, such as biuret methods, dye-binding techniques, and nephelometry are sensitive enough to accurately detect the low levels of protein present in CSF, but require destruction of a portion of the CSF sample, several reagents, and valuable preparation time [8]. Other common methods, such as UV-visible spectroscopy and other absorbance methods often require destruction of the sample by dilution, and in many cases are not appropriate due to significant error introduced by variability between different proteins [11] as well as the presence of various other molecules commonly found in biological fluids. Recent advances in the use of Raman spectroscopy suggest its use for CSF tubercular meningitis diagnostics, but early results indicate a method specificity of only 82%, with sensitivity also being an issue of concern for clinical translation [12]. Brillouin spectroscopy offers a possible method of quantifying total protein concentration without alteration of the fluid obtained from the body, representing a powerful simplification over most current techniques. As the protein concentration of a solution changes, so does the elasticity of the fluid [13], expressed as bulk modulus or compressibility. Brillouin spectroscopy is an optical technique which has been used extensively for material characterization and analysis [14–18]. The Brillouin frequency shift for a material is a function [19] of the material’s bulk modulus (B), density (ρ), and refractive index (n), as given by Eq. (1). sffiffiffiffi 2n B θ υ½Hz ¼ sin (1) λ ρ 2 sffiffiffiffi B represents the velocity of sound in the where ρ material [20] and θ ¼ π for backscattered light. In this way, the frequency shift of light represents a Doppler effect, as incoming light interacts with density fluctuations (sound waves) as they pass through the material [21]. By introducing proteins, which are less compressible than water, into a solution, the elastic properties of the solution change, as the Ćcompressibility of the solution approximates the

www.biophotonics-journal.org

409

weighted sum of its parts [13]. This paper reports experimental evidence that increased protein concentrations present in the CSF during bacterial meningitis can be measured by Brillouin spectroscopy, and suggests a possible approach to screening for the disease which is rapid, efficient, and nondestructive to the fluid sample. Finally, the sensitivity of the technique was found to increase vastly with fluid concentration by dialysis.

2. Results Our experimental results show that Raman spectroscopy and autofluorescence are insufficient to distinguish healthy and diseased samples. In order to assess the ability of Brillouin spectroscopy to screen for bacterial meningitis, solutions were created to mimic the constituents of CSF in both healthy and diseased patients. To model healthy fluid, porcine serum albumin and α-D-glucose were dissolved in PBS buffer, pH 7.4, in concentrations of 0.27 mg/mL and 0.7 mg/mL, respectively. Porcine serum albumin was chosen to represent the total protein in CSF due to its ready availability, similarity to human serum albumin [22], prevalence in meningococcal disease, and its globular structure representative of other globulins commonly found in spinal fluid samples [23]. Similarly, our diseased CSF model consisted of 2.5 mg/mL serum albumin with 0.3 mg/mL α-D-glucose, also in PBS buffer solution. It should be noted that the values for the diseased fluid model were taken as an average of cases of purulent and tubercular meningitis [23]. We first tested the Raman and autofluorescence spectra for our samples. The results are shown in Figure 1. We could observe no specific structures in Figure 1, except strong peaks at around 3400 cm–1 and weak peaks at ~1645 cm–1 (possibly from water) and ~350 cm–1 (resulting from the quartz container). No signature peaks were observed around the 2800– 3100 cm–1 region, indicating that the protein concentration (and also the glucose concentration) is below the detection threshold of the spontaneous Raman spectroscopy. No signature structures were presented, even within the region around 1652 cm–1, where the protein backbone vibration peak from serum albumin was expected. Therefore, we cannot distinguish the two samples solely based on Raman spectroscopy. In our test, the autofluorescence signal was shown as a smooth and broadband peak ranging from 300 cm–1 to 3000 cm–1. In this particular test, the “healthy” model showed higher autofluorescence than the “diseased” one. However, due to the impurities of the albumin sample, this result was not stable among our multiple tests, and we expect that it will be further complicated in clinical practice.

© 2014 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

410

Z. Steelman et al.: Brillouin screening for increased CSF protein

Figure 1 Illustration of the Raman spectrum of the “healthy” and “diseased” models. The inset is a magnified subset of the graph where the Raman peaks for protein would be expected. A reference spectrum of albumin can be readily found in the Ref. [24].

The numerical aperture of the objective lens does not drastically affect the Brillouin scattering results. We determined whether the numerical aperture of the objective lens could affect the results obtained. Theoretically, the Brillouin shift is determined by both the elastic properties of the sample and the scattering geometry (see Eq. (1)). Due to the nature of Gaussian beams, the wavefront of the focusing beam is curved except in the focusing plane. Therefore, additional variations in the scattering angle are introduced by these curved wavefronts, and the collected Brillouin peaks will be spectrally broadened (see Figure 2(a)). The use of high N.A. objectives may further complicate the problem, as the curvature of the wavefronts close to the focusing plane is larger than the setups with low N.A. lenses. Luckily,

this spectral-broadening is minimized by the backscattering geometry [25]. The equivalent pinhole introduced by the single-mode fiber is also capable of cleaning most of the photons with undesired scattering angles. To test these effects, we measured the linewidth of the Brillouin shift for DMSO using three different objectives (Thorlabs, RMS4X, N.A. = 0.10; Thorlabs RMS10X, N.A. = 0.25; Nikon CFI Plan Fluor, N.A. = 0.5). The images were obtained by an EMCCD. Fortunately, the linewidth was not significantly increased with the numerical aperture (see Figure 2(b)), indicating that the single-mode fiber could efficiently clean the scattered photons, leaving only the back-scattered photons (φ ≈ 180°). Determination of the Brillouin shifts for healthy and diseased CSF models. As previously discussed, model healthy and diseased fluids were prepared for spectroscopic characterization. To act as a control and a standard for normalization of results, PBS buffer alone was also analyzed. To diagnose any potential wavelength shift of the spectrometer beam source, three separate buffer control samples were tested at random intervals throughout the experiment. The Brillouin shift value of the PBS buffer was taken as an average of these tests, and used to correct for source wavelength drift during sampling. Empirically, we determined that the Brillouin shift increases with larger concentrations of glucose (sound velocity increasing with glucose concentration has previously been reported [26]). Proteins, as stated earlier, produce a similarly positive shift. This presented a potential challenge, as bacterial meningitis is often accompanied by a substantial (approx. 30–50%) decrease in CSF glucose levels [27], and concern was raised that any positive frequency shift obtained by increasing protein would be negatively compensated for by decreasing glucose. Upon pre-

Figure 2 (a) Illustration of the microscopic scattering geometry, here φ ≠ 180°; (b) The Brillouin shifts for DMSO obtained by three different objective lenses.

© 2014 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

www.biophotonics-journal.org

J. Biophotonics 8, No. 5 (2015)

411

Figure 3 A significant (P = 0.014) increase in the Brillouin shift of CSF is observed when protein and glucose concentrations mimic those present in fluid infected with bacterial meningitis. When the fluids are concentrated, the distinction is much greater. Box edges represent the middle two quartiles, and whiskers denote standard deviation.

paration and exposure of body-similar samples, however, it became immediately apparent that protein concentration has a significantly larger effect on the Brillouin shift than glucose, an effect which agrees with previously published data on the sound velocities of protein and glucose solutions [28]. In addition, it should also be noted that upon concentration of fluid protein by dialysis (performed in order to determine the sensitivity of the system to different protein levels), significant glucose loss should be expected in the sample, as glucose falls easily under the 2,000 molecular weight cutoff (MWCO) of the dialysis membranes used. Proteins, however, should be retained by the dialysis membrane, and so subjecting the fluid to dialysis has the effect of increasing the total protein concentration of the fluid, while decreasing the total glucose.

Upon measurement of the body-similar fluid samples, excellent distinction and separation between different samples was achieved. At protein and glucose concentration levels mimicking human spinal fluid during healthy and diseased states, the mean Brillouin shift of the “diseased” fluid (7.655 GHz) was clearly greater than that of the “healthy” fluid (7.616 GHz). In particular, a t-test proved with a high degree of confidence (P = 0.014) that the two samples were distinguishable based solely upon their Brillouin shifts. Predictably, the spectra from the fluid samples which had been concentrated by dialysis showed a much clearer distinction between the “healthy” and “diseased” samples (P = 5.71 × 10–5), with mean Brillouin shifts of 7.639 and 7.719 GHz, respectively, and almost no overlap between any data points in the sets. While the distinction between healthy model CSF and PBS was not significant (P = 0.215), diseased fluid was clearly distinguishable from the buffer control samples, again with a high degree of confidence (P = 1.35 × 10–4). As expected, increased fluid concentration of protein in general increased the Brillouin shift of the sample, with little regard to the glucose concentration. The results of this experiment, along with the high degree of confidence associated with this method, make it a potentially powerful diagnostic test for CSF total protein indicative of meningococcal inflammatory diseases. All data from these tests, including p-values calculated using a two-tailed, equal variance t-test, are summarized in Figure 3 and Table 1.

3. Discussion Brillouin spectroscopy provides a novel and nondestructive approach of screening for bacterial meningitis. It has substantially decreased the sample preparation time for spinal fluid acquired from in vivo through lumbar puncture. The extremely high level of confidence in the distinction between concentrated

Table 1 p-values for relevant sample comparisons. Two samples are considered significantly distinguishable by Brillouin spectroscopy if p < 0.05. SD = Standard Deviation. Sample Mean ± SD (GHz)

“Healthy” 7.616 ± 0.028

“Diseased” 7.655 ± 0.033

Buffer 7.601 ± 0.033

Conc. “Diseased” 7.719 ± 0.027

“Healthy” 7.616 ± 0.028 “Diseased” 7.655 ± 0.033 Buffer 7.601 ± 0.033 Conc. “Healthy” 7.639 ± 0.037



0.014

0.215



0.014



1.35 × 10–4



0.215

1.35 × 10–4











5.71 × 10–5

www.biophotonics-journal.org

© 2014 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

412

Z. Steelman et al.: Brillouin screening for increased CSF protein

“healthy” and “diseased” fluid is encouraging, as it represents the potentially vast improvement that a small enhancement of system sensitivity could bring. Even without concentration of the fluid by dialysis, our results are indicative of a screening procedure in which alteration of the fluid as taken from the body is nonessential, and which is rapid, safe, and reliable. These results are significant, in that they offer a method of detecting elevated CSF protein levels without preventing a physician’s ability to perform other valuable diagnostic tests on the sample. A protein content of approximately 2.5 mg/mL is typical for bacterial meningitis cases. In reality, however, a range of increased protein levels exist. While this study only demonstrates the viability of Brillouin spectroscopy as a diagnostic tool, further analysis should be performed to determine detection limits of the technique. In addition, the presence of an elevated white blood cell count (while likely insignificant to the bulk fluid properties) on the observed Brillouin shift should be investigated as well. The relatively strong elastic scattering is one of the major limiting factors for our measurements. The elastic peaks and the related parasitic diffraction become the main sources of contamination in determining the Brillouin shift. Recent developments in background-free Brillouin spectroscopy efficiently remove the elastic peaks by inserting atomic/molecular absorption cells into the beam path [29]. This technique may be beneficial to improving the signal quality in this study. The application of Brillouin spectroscopy requires strict spectral stabilization of the laser emission. The solid state laser source employed in this study showed relatively good short- and long-term spectral stabilities. However, spectral drift, although minute, still occurred over the course of the experiment (nonlinear wavelength drift with time most likely caused an uneven distribution of measurements as seen in the healthy and diseased samples). For future routine use, diode laser sources are preferred, due to their superior tunabilities and stabilities. Additional frequency locking procedures should be performed for further stabilization [30]. While we showed that Raman spectroscopy alone was incapable of distinguishing the samples, Raman spectroscopy could still offer substantial information concerning analytes within the fluid in some circumstances. For instance, analysis of single bacterial cells during bacterial meningitis using Raman spectroscopy has shown promise as an analysis technique [31]. In addition, the comparison of simultaneous Brillouin and Raman data as a means of tissue analysis has recently been demonstrated as a valuable research tool, with a strong potential for diagnostic applications [32]. The results of this study suggest that small improvements in design leading to improved spectral

© 2014 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

dispersion and spatial resolution have the potential to develop this technology into an important and powerful diagnostic tool. Improvements in laser linewidth and stability are also needed to improve the reliability and accuracy of this method. In general, Brillouin spectroscopy has shown to be a promising and unique diagnostic tool, which could provide physicians and pathologists with additional needed information to confirm diagnoses and save lives.

4. Methods A single-mode, solid-state laser (Lasermate Group, Inc.; model: GMSL-532-100FHA) with a calibrated center wavelength of 531.9587 nm ± 0.3 pm was utilized as the incident pump source for both Raman and Brillouin spectroscopies. Its nominal output linewidth was approximately 640 kHz, with a maximum output power of ~100 mW. During the Raman tests, we adopted a simple back-scattering geometry for our setup. The Raman scattered photons were directed into a spectroscope (Shamrock 303, Andor Technology Ltd.), dispersed by a 1200 line/mm reflective grating, and collected by a EMCCD camera (Newton 971, Andor Technology Ltd., cooled at –50 °C). The pump power was ~50 mW. The integration time for the CCD camera was 50 s. When measuring the Brillouin spectra, we employed a two-stage VIPA (virtually imaged phased array) spectrometer setup. An optical isolator (Electro-Optics Technology, Inc., Model: BB-8-05-I-090) was placed in the beam path to avoid unwanted feedback from stray reflections and scattering. The beam was then passed through a 50–50 non-polarizing beamsplitter, before being focused by a microscope objective lens (Nikon, Inc., CFI Plan Fluor 20×, N.A. = 0.5). Samples were placed in quartz cuvettes (Starna Cells Inc.), and backscattered light was collected back through the objective lens, and directed into a single-mode fiber (Fibercore Inc., model: SM600, 1 meter) to the 2-stage VIPA spectrometer, which mirrored the design employed by Scarcelli and Yun [33]. Specifically, the VIPAs (Light Machinery Inc., model: OP-5642) have a free spectral range of 33.3 GHz, and were designed explicitly for a wavelength of 532 nm such that one surface reflects 99.5% of the light at 532 nm, and the other – 95%. To minimize optical diffraction, 200 lenses were used throughout the spectrometer, and a 1000 mm focal length lens placed after the second VIPA allowed for the Brillouin scattered components to be sufficiently separated from the fundamental radiation. The resulting spectrum was acquired on a CCD camera (Moravian Instruments, model: G2-8300, pixel size 5.4 μm × 5.4 μm for sample measurements;

www.biophotonics-journal.org

J. Biophotonics 8, No. 5 (2015)

Figure 4 (a) Schematic diagram of a Brillouin spectrometer. (b) A more detailed schematic of a two-stage VIPA spectrometer (LF: line filter at 532 nm, CL: cylindrical lens, SL: spherical lens).

Andor Inc., Model: Newton 970, pixel size: 16 μm × 16 μm for linewidth measurements). A typical data acquisition time was 20 s. The complete spectrometer is shown in Figure 4. Samples consisted of porcine serum albumin (Sigma-Aldrich, purity ≥98%) in phosphate-buffered saline (Sigma-Aldrich, 0.01 M, pH 7.4), as well as αD-Glucose (Sigma-Aldrich, purity ≥96%) in concentrations made to imitate in vivo CSF total protein and glucose concentrations of healthy and diseased patients. Serum albumin was substituted for total protein due to both its globular structure [22] and prominence in diseased spinal fluid compositions. A third sample consisting of only PBS buffer acted as a control. Since Brillouin frequency shifts strongly depend upon the temperature of the sample [34], all three solutions were placed in the room housing the spectrometer for several hours prior to data acquisition to thermally equilibrate the samples with the environment, and were then placed in quartz cuvettes. In order to determine the sensitivity of the screening method to larger protein concentrations, a small amount of both body-similar samples were also placed in dialysis cassettes (Slide-A-Lyzer, 2K MWCO, 3 mL), which were immersed in a high molecular weight poly(ethylene glycol), MW = 3350, and allowed to dialyze for 100 minutes. This allowed the water and glucose in the solution to diffuse across the membrane to the highly hygroscopic poly (ethylene glycol), while the protein remained within and the solution became concentrated. Measurement of the total fluid volume after dialysis indicated that roughly 50% of the water in the solution crossed the membrane in all cases, leaving behind a solution of approximately twice the concentration of serum al-

www.biophotonics-journal.org

413

bumin (glucose has a molecular weight of less than 200 Da, and was thus free to cross the membrane). These concentrated solutions were allowed to equilibrate thermally as well, and were then tested using the spectrometer. To correct for any thermal drift of the source wavelength, three groups of control samples of PBS were taken at random intervals between samples intended for screening, and a linear correction was made to all data points based on the best fit line of the control samples by time of data acquisition. For each sample, ten consecutive measurements were taken (including 30 total PBS control measurements). A typical spectral image is shown in the abstract figure, where the Brillouin peaks (Stokes and antiStokes) are clearly visible. The raw images were processed using a developed MATLAB code which calculated the distance between the weighted centers of two corresponding Brillouin peaks after removing noise and correcting for a slow laser drift. First, the algorithm calculates the average intensity of a background region of the output image, and then normalizes each image’s overall intensity to its background to ensure accurate processing between data sets. The algorithm then effectively scans the image around the Brillouin peaks with a small window (approximately 60 μm by 60 μm), and, upon finding the most intense area, expands into a circular region and calculates the center of that region, weighted for intensity. In this way, any background noise (which has been observed to be as high as 75% of the Brillouin peak intensity) does not affect the center of the Brillouin peak as it would if we simply predefined a window in which to search for Brillouin spots. The linear distance between two corresponding Brillouin spots was then normalized to the drift in peak distance found in PBS control samples. The drift was found to be linear in time, and was corrected based upon the exact time of exposure. The Brillouin peak was only slightly more intense than the background, while the central peak was orders of magnitude more intense, so special efforts were undertaken to eliminate noise for the camera, which was placed in an opaque black box, along with the VIPA optics and related lenses. Significant data exists in the literature regarding the speed of sound in water, as well as the refractive index of water, in relation to temperature and salinity [35–38]. In order to calculate an accurate Brillouin frequency shift, the average peak-to-peak distance of PBS buffer calculated above was normalized to the expected Brillouin shift of water, using identical values for salinity and room temperature. As all other components make up less than 10% of the total dissolved solids, the NaCl concentration of the PBS solution was used for salinity in the calculation. An empirically derived expression for the speed of sound in water [39] (V) based on its tem-

© 2014 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

414

Z. Steelman et al.: Brillouin screening for increased CSF protein

perature (T) in Celsius, and salinity (S) in parts per thousands (ppt) is given by Eq. (2). V ¼ 1449 þ 4:6T  0:055T 2 þ 0:003T 3 þ ðS  35Þ ð1:39  0:012TÞ

(2)

Using Eqs. (1) and (2), and an approximate refractive index of 1.334 [38], the average peak-to peak distance of PBS buffer control samples was normalized to a calculated value of 7.6007 GHz. In order to maintain consistency, the arc length between adjacent elastic peaks (traced along the interference line using spline interpolation) was normalized to the free spectral range of the VIPA etalons (33.3 GHz) to determine a pixel/ GHz conversion ratio for each image. Finally, deviations from the average control peak-to-peak distance were converted to GHz shift values using the calculated conversion ratios, and a Student’s t-test was performed between each sample’s data set to determine a confidence level for the screening method. Results were termed significant if the p-value obtained from the t-test was less than α = 0.05. Author contribution statement ZS developed the experiment, created the MATLAB code for data analysis, created Figure 3, and authored the manuscript. ZM designed and built the setup, performed the N.A. experiment, and created Figures 1, 2 and 4. AJT contributed to the data analysis and performed the statistical analysis. VVY provided guidance and funding. All authors reviewed the manuscript. Acknowledgements This work was partially supported by National Science Foundation Grants ECCS-1250360, DBI-1250361, and CBET-1250363. Author biographies online.

Please see Supporting Information

References [1] H. El Bashir, M. Laundy, and R. Booy, Arch. Dis. Child., 88, 615–620 (2003). [2] J. Mongelluzzo, Z. Mohamad, T. R. Ten Have, and S. S. Shah, JAMA, 299, 2048–2055 (2008). [3] M. N. Swartz, M. D., N. Engl. J. Med. 351, 1826–1828 (2004). [4] H. Pfister, W. Feiden, and K. Einhäupl, Arch. Neurol. 50, 575–581 (1993). [5] G. E. Enwere and S. K. Obaro, Lancet 358, 1549 (2001). [6] A. R. Tunkel and W. M. Scheld, Curr. Infect. Dis. Rep. 4, 7–16 (2002). [7] K. Stuertz, I. Merx, H. Eiffert, E. Schmutzhard, M. Mäder, and R. Nau, J. Clin. Microbiol. 36, 2346–2348 (1998). [8] B. Karlsson and C. Alling, Clin. Chim. Acta 105, 65– 73 (1980).

© 2014 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

[9] C. A. Pennock, L. P. Passant, and F. G. Bolton, J. Clin. Pathol. 21, 518–520 (1968). [10] M. C. Brouwer, G. E. Thwaites, A. R. Tunkel, and D. van de Beek, Lancet 380, 1684–1692 (2012). [11] S. Metsämuuronen, M. Mänttäri, and M. Nyström, Desalination 283, 156–164 (2011). [12] R. Sathyavathi, N. C. Dingari, I. Barman, P. S. R. Prasad, S. Prabhakar, D. Narayana Rao et al., J. Biophotonics 6, 567–572 (2013). [13] S. H. Wang, L. P. Lee, and J. S. Lee, J. Acoust. Soc. Am. 109, 390–396 (2001). [14] J. D. Bass, J. Geophys. Res. 94, 7621–7628 (1989). [15] R. E. Gagnon, H. Kiefte, M. J. Clouter, and E. Whalley, J. Chem. Phys. 92, 1909 (1990). [16] P. H. Gammon, H. Kiefte, and M. J. Clouter, J. Phys. Chem. 87, 4025–4029 (1983). [17] J. M. Jackson, S. V. Sinogeikin, and J. D. Bass, Am. Mineral. 85, 296–303 (2000). [18] G. Scarcelli, S. Kling, E. Quijano, R. Pineda, S. Marcos, and S. H. Yun, Invest. Ophthalmol. Vis. Sci. 54, 1418–1425 (2013). [19] Y. H. Ko, K. J. Kim, and J. H. Ko, J. Korean. Phys. Soc. 63, 2358–2361 (2013). [20] M. Habrioux, S. V. D. Freitas, J. A. P. Coutinho, and J. L. Daridon, J. Chem. Eng. Data 58, 3392–3398 (2013). [21] Y. Zhang, G. Fu, Y. Liu, W. Bi, and D. Li, Optik 124, 718–721 (2013). [22] K. Monkos, Biochim. Biophys. Acta. 1748, 100–109 (2005). [23] C. J. Brackenridge, J. Clin. Pathol. 15, 206–210 (1962). [24] A. Saha and V. V. Yakovlev, J. Biophotonics 3, 670– 677 (2010). [25] G. Antonacci, M. R. Foreman, C. Paterson, and P. Török, App. Phys. Lett., 103, 221105 (2013). [26] D. E. Smith and W. C. Winder, J. Food Sci. 48, 1822– 1825 (1983). [27] P. R. Donald, C. Malan, and A. van der Walt, J. Pediatr., 103, 413–415 (1983). [28] A. K. Nain, R. Pal, and R. K. Sharma, J. Chem. Thermodyn. 43, 603–612 (2011). [29] Z. Meng, A. J. Traverso, and V. V. Yakovlev, Opt. Express 22, 5410–5415 (2014). [30] C. E. Wieman and L. Hollberg, Rev. Sci. Instrum. 62, 1–20 (1991). [31] M. Harz, M. Kiehntopf, S. Stöckel, P. Rösch, E. Straube, T.Deufel et al.,J. Biophotonics 2,70–80 (2009). [32] F. Palombo, M. Madami, N. Stone, and D. Fioretto, Analyst 139, 729–733 (2014). [33] G. Scarcelli and S. H. Yun, Opt. Express 19, 10913 (2011). [34] H. Gu, H. Dong, G. Zhang, J. He, and H. Pan, IEEE Sens. J. 13, 864–869 (2013). [35] N. Bilaniuk and G. S. K. Wong, J. Acoust. Soc. Am. 93, 1609–1612 (1993). [36] J. C. Carman, J. Acoust. Soc. Am. 131, 2455–2458 (2012). [37] B. Richerzhagen, Appl. Opt. 35, 1650–1653 (1996). [38] J. V. Leyendekkers, Mar. Chem. 5, 29–42 (1977). [39] J. G. Hirschberg, J. D. Byrne, A. W. Wouters, and G. C. Boynton, Appl. Opt. 23, 2624–2628 (1984).

www.biophotonics-journal.org

Brillouin spectroscopy as a new method of screening for increased CSF total protein during bacterial meningitis.

Bacterial meningitis is a disease of pronounced clinical significance, especially in the developing world. Immediate treatment with antibiotics is ess...
197KB Sizes 0 Downloads 3 Views