Open Access
Add to BibSonomyAbstract
To facilitate the application of plasmonic nanoparticles (PNPs) in high-throughput detection, we develop a hyperspectral imaging system (HSIS) combining dark-filed microscopy and imaging Fourier transform spectrometry to measure scattering spectra from immobilized PNPs. The current setup has acquisition time of 5 seconds and spectral resolution of 21.4 nm at 532.1 nm. We demonstrate the applicability of the HSIS in conjunction with spectral data analysis to quantify multiple types of PNPs and detect small changes in localized surface plasmon resonance wavelengths of PNPs due to changes in the environmental refractive index.
©2011 Optical Society of America
1. Introduction
Plasmonic nanoparticles (PNPs) have been increasingly used as optical tags for sensitive detection of biomolecules such as DNA [1–3] and proteins [4–6]. Localized surface plasmon resonance (LSPR) of a PNP enhances its scattering cross section equivalent to more than 105 organic fluorescent dye molecules in their abilities to generate optical signals [7]. The strong signal from PNPs has enabled detection of single target molecules with common and inexpensive instruments such as a dark-field microscope [3–7]. Compared to fluorescent dyes, PNPs are not susceptible to blinking, photobleaching and quenching, making them more suitable for quantitative measurements over extended periods of time.
PNPs have also been used as optical sensors in biomedical research to monitor biological interactions which cause changes in the refractive index near or on the surface of a PNP. The resultant shifts in the LSPR wavelength of a PNP can be determined experimentally by spectroscopic measurements of light scattering or extinction. Inter-particle surface plasmon coupling also affects LSPR and can be employed as a molecular ruler [8] or an indicator of interaction between two biomolecules [9]. Recent developments of LSPR as biosensors are reviewed in [10]. With the development of hyperspectral imaging techniques, the strong optical signals and small sizes of PNPs represent great opportunities for simultaneous molecular imaging and local environmental sensing in live cells [9,11].
To achieve high-throughput biomolecular detection using PNPs in microarrays, acquisition speed for a large field of view needs to be maximized. Multicolor detection capability is desirable to increase the throughput and provide proper controls in bioassays [2]. Therefore, improving acquisition speed of instruments for simultaneously measuring spatial and spectral information facilitates the application of PNPs in high-throughput detection. Measuring scattering spectra of single PNPs with high spectral resolution is commonly achieved by using grating-based spectrographs. Using two-dimensional (2D) array detectors enables acquiring spectra from pixels along a line field of view with an exposure and 2D hyperspectral images can be obtained by push-broom scanning of the sample [12]. Becker et al. proposed using an electronically addressable mask to selectively measure spectra from up to 20 PNPs per exposure for situations where only a few PNPs are present within the line field of view [13]. This method is not suitable for high-resolution imaging. Liu et al. achieved high-resolution imaging of a large field of view by using a grating monochromator to select a 2 nm spectral band for dark-field illumination, scanning the band and capturing narrow-band images sequentially [14]. Hyperspectral images of Au PNPs and nanowires were obtained with a relatively high spectral resolution and an exposure time of 2 seconds per band.
We took a different strategy to improve the acquisition speed of hyperspectral imaging systems by reducing the spectral resolution or the number of resolved spectral bands. Although conceptually simple, implementation of the strategy is not trivial because the effects of reduced spectral resolution need to be monitored and assured not to impede the quantitative detection ability of PNPs. We constructed a hyperspectral imaging system (HSIS) based on imaging Fourier transform spectrometry. Hyperspectral images were obtained by changing optical path-length difference (OPD) between two beams in an interferometer, recording the interference intensity as a function of the OPD (interferogram), and Fourier transforming the interferograms pixel by pixel. We chose to use imaging Fourier transform spectrometry for this study mainly because of its flexibility to vary spectral resolution without affecting spatial resolution by simply selecting a different OPD range, eliminating the need to change the optical setup. We report the results of characterizing the HSIS and demonstrate the usefulness of the proposed method to achieve quantitative measurements of PNPs as multicolor optical tags and sensors for local refractive index.
2. Imaging Fourier transform spectrometry based hyperspectral imaging system (HSIS)
The HSIS mainly consists of two parts: a transmission dark-field microscope and a Michelson interferometer (Fig. 1(a) ). Light from a Xenon arc lamp (66452, Newport) passes a longpass filter with a cutting-on wavelength of 400 nm and illuminates the sample through a dry dark-field condenser (NA 0.92~0.8, Olympus). Light scattered off PNPs is collected by a water-immersion objective (20X, NA 0.5, Olympus) and goes into an imaging Fourier transform spectrometer based on a Michelson interferometer. A cube beam-splitter (CSMH, Sigma Koki) splits the beam from the microscope into two beams. One is reflected toward a fixed mirror, M2, and the other is transmitted toward a movable mirror, M1, which is attached to a piezoelectric translator (PZT; P-841.20, Physik Instrumente). At the exit port of the Michelson interferometer, an achromatic doublet lens (NT32-327, Edmund Optics) focuses the reflected beams from M1 and M2 onto the surface of a charge-coupled device (CCD; EC1380, Prosilica), forming an image of the sample under the dark-field microscope. The field of view is 659 μm × 496 μm and the lateral resolution is 1.32 μm with the components used. The lateral resolution is measured as the 10%-90% distance of the edge response from a 1951 USAF resolution target.
Fig. 1 (a) Schematic diagram of the imaging Fourier transform spectrometer consisted of a dark-field microscope and a Michelson interferometer. M1, M2, M3: mirrors, BS: beam-splitter; CCD: charge-coupled device camera. (b) Example interferogram of a 532.1nm laser. (c) Spectrum of the laser is obtained by taking Fourier transform of the interferogram. The spectral resolution, Δk, is measured from the FWHM of the laser spectrum. (d) Spectral resolution of the HSIS at 532.1 nm versus OPD range. The dashed line and red squares represent theoretical and experimental values, respectively.
The OPD between the two beams in the Michelson interferometer is varied in steps by driving the PZT and the CCD sequentially captures an image at each OPD. Operations of the PZT and the CCD are automated by using a program written in LabVIEW (National Instruments). In order to cover the visible spectral range of 400-800 nm, the interferograms are sampled at 200 nm OPD intervals, corresponding to a step size d = 100 nm for the PZT. Theoretical spectral resolution of the HSIS, Δk, is dependent on the total OPD range, L, by this equation Δk = 1.21/L (μm−1). To characterize the spectral resolution, we recorded interferograms from a 532.1nm laser and performed Fourier transform pixel by pixel. Figures 1(b) and 1(c) shows an exemplary interferogram recorded at one pixel and the corresponding spectrum, respectively. Note the x-axis of the spectrum is in wavenumber (1/cm) or reciprocal wavelength to demonstrate the Fourier transform relation between the interferogram and the spectrum. Figure 1(d) shows that the theoretical spectral resolutions of the HSIS at various OPD ranges match well with those measured from the full width at half maximum (FWHM). The error bars indicate standard deviations of spectral resolution measured from 10 randomly selected pixels in the field of view. All error bars in Fig. 1(d) are within 2% of the corresponding mean values.
3. Calibration of OPD step size at different locations in an image
Due to slight misalignment of optical components in the setup and off-axis effects, the step size in OPD varies across the pixels in a captured image. The step size represents the sampling interval of the interferogram and needs to be determined accurately in order to calculate the correct scale of wavelength in the resultant spectrum. We assume that the step size in OPD for an on-axis pixel is 2d and that for an off-axis pixel is 2d*cos(ϕ) where ϕ is the incidence angle to the mirrors in the interferometer as shown in Fig. 2(a) . In practice, it is difficult to directly determine the position and orientation of the optical axis and the corresponding angle ϕ for every pixel. To practically obtain the OPD step sizes for all the pixels in an image, we used a bandpass filter as a sample under bright-field illumination and recorded interferograms by the HSIS. We measured a spectrum of the bandpass filter under the same illumination condition with a commercial imaging spectrograph which consisted of a monochromator (SP-2150, Princeton Instruments) and a back-illuminated EMCCD (Newton DU-970N, Andor Technology). To determine the optimal value of the OPD step size, we calculated cross-correlations between the accurate spectrum measured by the commercial imaging spectrograph and those obtained from the interferograms using various OPD sampling intervals. By repeating this procedure for each of the pixels in the field of view the optimal OPD step size for each pixel was obtained and stored for later use. Figure 2(b) shows exemplary calibrated spectra measured from four pixels which demonstrate similarities to that measured by the commercial imaging spectrograph. To evaluate the accuracy of wavelength calibration, we utilized the calibrated OPD step sizes to calculate the spectra measured from the 532.1 nm laser. The average and standard deviation of the peak wavelength among 10 randomly selected pixels in the field of view was 531.78 ± 0.33 nm. For comparison, the average peak wavelength was 534.57 ± 1.09 nm when a constant OPD step size of 100 nm was used.
Fig. 2 (a) Geometry of an off-axis ray shows its OPD step size of 2dcos(ϕ). (b) Spectrum of the light source through a bandpass filter measured by the commercial imaging spectrograph (black) and those at four pixels in a hyperspectral image measured by the HSIS. The spectra from the upper-left, upper-right, bottom-left and bottom-right corners are plotted in red, green, blue and pink lines, respectively.
4. Experimental results of measuring scattering spectra of PNPs by the HSIS
To validate the HSIS for the intended use, we measured scattering spectra of three types of PNPs including 40 nm gold, 40 nm silver and 80 nm silver nanoparticles (BBI International, Inc.) and compared the results to those measured by the commercial imaging spectrograph. PNPs were coated on glass slides which were pre-cleaned in boiling piranha solution (3:1 concentration of H2SO4:H2O2) for one hour and modified with 2% aminopropyltriethoxysilane for 20 minutes. Because of the relatively wide FWHM of scattering spectra from the PNPs, the requirement for spectral resolution can be loosened and the OPD scanning range can be reduced substantially to speed up data acquisition. In the following PNPs measurements, we used an OPD range of 16 μm with a step size of 200 nm, corresponding to 80 images captured for one hyperspectral imaging data cube. The spectral resolution is about 21.4 nm (Fig. 1(b)). Based on our experience, this OPD range achieves a reasonable trade-off between spectral resolution and acquisition time because it covers most oscillations in the interferograms measured from the PNPs. Prior to taking Fourier transform of the interferograms, we performed apodization with a triangular function and zero-padding to reduce the effect of a finite window width and help smooth out the spectra, respectively.
Figures 3(a) –3(c) show dark-field images of the three types of PNPs captured by a color CCD (Micro Publisher 3.3RTV, Q-Imaging). Scattering spectra of these PNPs measured by the commercial imaging spectrograph and the HSIS are shown in Fig. 3(d) and 3(e), respectively. We note that the spectra measured by the two instruments were not from the same PNPs. Instead, scattering spectra of 5-20 PNPs were averaged to account for inter-particle variations. Although the spectra measured by the HSIS show a smoothing effect due to a lower spectral resolution, the peak wavelengths and trends in intensity match well with those of spectra measured by the commercial imaging spectrograph. After correcting the effects of non-uniform spectral characteristics of the optical components in both systems, we obtained scattering spectra of single 40 nm gold nanoparticles as shown in Fig. 3(f). LSPR peak wavelength of the 40 nm gold nanoparticles measured by the commercial imaging spectrograph and the HSIS was 546.4 nm and 544.7 nm, respectively. These values do not agree with the theoretical LSPR peak wavelength of 537 nm predicted by Mie theory. Possible causes for the discrepancies include heterogeneities in size and shape of the PNPs as well as effect of the glass substrate in the vicinity of the PNPs [15,16]. Intensities of the HSIS spectrum at short wavelengths deviate considerably from the theoretical values, which could be attributed to lower signal-to-noise ratios (SNRs) at shorter wavelengths due to lower quantum efficiency of the CCD and higher influence of environmental vibrations. Lower signals detected at shorter wavelengths result in larger multiplicative correction factors used for correcting the system spectral response. In order to avoid amplifying the noise at shorter wavelengths, spectral analyses described in the following section were applied to raw spectral data.
Fig. 3 Color images of (a) 40 nm gold (b) 40 nm silver and (c) 80 nm silver. Average scattering spectra of 40 nm gold (green lines), 40 nm silver (blue lines) and 80 nm silver (red lines) nanoparticles measured by (d) the commercial imaging spectrograph and (e) the HSIS. (f) Scattering spectra of a single 40 nm gold nanoparticle predicted by Mie theory (blue line) and measured by the commercial imaging spectrograph (red line) and the HSIS (green line).
5. Applications of combining the HSIS with spectral analysis on PNPs detection
To demonstrate the applicability of using the HSIS to quantify multiple types of PNPs in a microarray format, we mixed 40 nm silver and 40 nm gold nanoparticles with 5 different concentration ratios between 0.02 and 0.1 and incubated the mixed PNPs on glass slides for 10 minutes. The slides were prepared as in the single PNPs experiment. We acquired a hyperspectral imaging data cube in 5 seconds with an exposure time of 50 ms per image. To facilitate visualizing hyperspectral images, a false-colored RGB image of immobilized PNPs was obtained by setting the values of the red, green and blue channels proportional to the integrated intensities in wavelength ranges of 600-700 nm, 500-600 nm, and 400-500 nm, respectively. Appropriate balance between the three color channels was achieved by visual comparison between acquired color images as those shown in Figs. 3(a)–3(c) and false-color images generated from the corresponding spectra. As shown in Fig. 4(a) , the surface density of the PNPs is fairly uniform. For quantitative spectral analysis, intensities of scattering spectra of all the pixels in the same field of view were summed together after background subtraction (Fig. 4(b)). To simplify the problem of quantifying the contribution of each type of PNPs to the mixed spectra, we assumed that there was negligible coupling of surface plasma between PNPs so that the intensity and spectral pattern of one nanoparticle were insensitive to the presence of other nanoparticles. The measured mixed spectra, represented by a row vector m, can be expressed as a linear combination of pure-component spectra by m = cS where c is a row vector containing weights of the pure components and each row in the matrix S contains the spectrum of a pure component. We took the average scattering spectra of single PNPs as the pure component spectra and obtained the weights by using ordinary least squares regression. Pixels with peak intensity above a certain threshold were assumed to be contaminates or large aggregates on the slides and excluded from the analysis. We calculated ratios of extracted weights between silver and gold PNPs to serve as a quantifiable outcome in accordance with typical DNA microarray assays in which proper calibration and/or controls are used for quantitative measurements. As shown in Fig. 4(c), the ratios of extracted weights averaged from three different fields of view correlate linearly with the original concentration ratios of PNPs in the mixed solutions. Since the intensity of aggregated PNPs has a quadratic dependence to the number of nanoparticles [9], there is an upper limit in surface density of the PNPs for which the linear regression method works. Based on the high linearity between the original concentrations and the extracted weights, we validated the assumption of negligible inter-PNPs coupling for the surface densities of PNPs used in this experiment.
Fig. 4 (a) False-color image of mixed PNPs immobilized on a slide. Scale bar is 100 μm. (b) Spectra of mixed PNPs with different concentration ratios and corresponding fitting results (dashed lines). (c) Extracted weights of silver and gold PNPs on glass surface show a linear correlation between silver/gold intensity ratios and concentration ratios. Error bars represent the standard deviation among three different fields of views.
The ordinary least squares method we used to extract the weights of individual type of PNPs from mixed spectra requires predetermined pure component spectra. Since the shape of scattering spectrum is very sensitive to the local environment of the PNPs, pure component spectra present in real samples may deviate from those measured from single PNPs. The issue can be mitigated by a multivariate curve resolution algorithm which uses constrained alternating least squares and has been demonstrated to quantify sample concentration and resolve overlapping fluorescence emission spectra in gene expression data [17,18].
To demonstrate the application of the HSIS to detect shifts in LSPR peak wavelength due to changes in the local environment of PNPs, we measured scattering spectra of individual 40 nm silver nanoparticles immobilized on glass surface and immersed in solutions with various refractive indices. We also used an exposure time of 50 ms per image. Water and glycerol were mixed in various proportions to prepare solutions with refractive indices between 1.333 and 1.475 in increments of 0.028. We define center-of-mass wavelength as an indicator of changes in spectral profile [19] in response to changes in the environment of individual PNPs.
where λ is wavelength and I is the intensity at the corresponding wavelength. The wavelength range for the spectral center-of-mass analysis was chosen to be the union of the FWHMs of scattering spectra measured from the 40 nm silver nanoparticles in pure water and pure glycerol. Figure 5(a) shows the scattering spectra of one silver nanoparticle in different refractive indices and Fig. 5(b) shows the corresponding center-of-mass wavelengths. The scattering spectra of the silver nanoparticle shown in Fig. 5(a) are summations of spectra measured from 3 × 3 or 4 × 4 pixels. At this surface density, the fraction of aggregates is well below 1% [13]. As shown in Fig. 5(c), a high degree of linearity (R2 = 0.966) exists between the change in environmental refractive index and the average shift in center-of-mass wavelengths among 20 PNPs randomly selected from a total of about 140 PNPs in a field of view. In addition to noise in the measured spectra, heterogeneities in size and shape of the silver nanoparticles also contribute to variations in peak wavelength and sensitivity to environmental refractive index between PNPs. Despite moderate spectral resolution in the range of 12-49 nm, we successfully demonstrate the detection of shifts in LSPR wavelength of 40 nm silver PNPs corresponding to a change in environmental refractive index as small as 0.028.Fig. 5 (a) Scattering spectra of one silver nanoparticle immersed in solutions with different refractive indices. (b) Center-of-mass wavelengths calculated from (a) correlate linearly with the refractive indices of the solutions. (c) The average shift in spectral center of mass correlates linearly with changes in refractive index of the medium. The error bars are standard deviations among 20 nanoparticles.
6. Discussion and summary
The major feature of the HSIS presented in this paper is fast acquisition of microscopic hyperspectral images with extended field of views at the expense of spectral resolution. The current acquisition time of the HSIS (5 seconds) is at least one order of magnitude shorter than previously reported practical implementations of fast hyperspectral imaging systems for measuring scattering spectra of PNPs [13,14]. Scattering spectra of 160 particles in a field of view of 200 μm × 200 μm could be acquired with a spectral resolution of 1.6 nm in 200 seconds by a spectrograph-based system [13]. A narrow-band wide-field system achieved acquisition time of 300 seconds with a spectral resolution of 4 nm in the wavelength range of 500-800 nm [14]. Although the spectral resolution of the HSIS is about one order of magnitude lower than spectrograph-based systems, we point out that the spectral resolution needed for a particular application is usually not well studied and optimized because the high spectral resolution provided by imaging spectrographs is more than adequate. By sacrificing the often abundant spectral information we improved both acquisition speed and field of view, which are advantageous for imaging large populations of living cells. The sensitivity of the HSIS to changes in environmental refractive index, which is about ~0.028, could be useful in a number of applications. For examples, Curry et al. used peak wavelength shifts in scattering spectra of gold PNPs to discriminate between receptor mediated uptake and non-specific uptake of nanoparticles in living cells and reported that the receptor-mediated internalization of antibody-coated PNPs resulted in a change in the refractive index from 1.44 to 1.53 [20]; Aaron et al. used aggregation of antibody-conjugated Au PNPs as a contrast enhancing agent to image over-expression of the epidermal growth factor receptor and measured a ~100 nm shift in LSPR peak wavelength in living cancer cells [9]. The HSIS has sufficiently high spectral resolution to detect the wavelength shifts in these biomedical applications.
The acquisition time of measuring scattering spectra of PNPs from an extended field of view is mainly limited by total exposure time due to the low level of scattered light. The SNR and hence the acquisition speed of the reported HSIS can be further improved by using more intense illumination irradiance, an objective lens with higher numerical aperture and a detector with higher quantum efficiency. Using a frame-transfer camera can reduce the camera read-out time to a negligible level, which could be a significant improvement for a large number of pixels per frame. A common-path Sagnac interferometer can be used in place of the Michelson interferometer to reduce the effects of environmental vibrations for applications where a proper vibration-isolation mechanism is not applicable. The field of view of the current setup is limited to a circular area which is non-uniformly illuminated by the dark-field condenser (Fig. 4(a)). One possible method to address this issue is using a planar waveguide to provide more uniform dark-field illumination by evanescence wave [2].
Although the proposed strategy of improving the acquisition speed is demonstrated with a Fourier transform spectrometry-based hyperspectral imaging system, it is also applicable to other hyperspectral imaging techniques. Since instrumental factors capable of improving the SNR can be effectively applied to shorten the acquisition time of a system, we focus our discussion on the improvement factor that could be achieved in the same system by only changing the spectral resolution. For narrow-band scanning systems, increasing the bandwidth by a factor of M results in a reduction in the total exposure time by a factor of M2. The exposure time for each pass-band can be decreased by a factor of M to achieve a similar SNR, and the total number of bands also decreases by a factor of M. For snapshot hyperspectral imaging systems [21,22], broadening the bandwidth by M-fold results in an M-fold increase in the pixel count per frame since the total number of elements in a hyperspectral imaging data cube is constant.
A related and noteworthy technique widely used in biosensors for detecting minute refractive index changes is surface plasmon resonance at the interface between a dielectric medium and a gold or silver thin film. Wide-field surface plasmon resonance imaging can be achieved by illuminating the thin film with a collimated monochromatic beam at or near the resonance angle and measuring reflected intensities as an image. The lateral resolution in surface plasmon resonance imaging is limited to several microns due to the propagation length of surface plasmon waves [23]. Although more localized excitation and wide-angle detection of surface plasmon waves have been demonstrated by using high numerical aperture objective lenses, improvements in spatial resolution are accompanied by decreased sensitivities [24]. The current HSIS setup could be readily modified to achieve diffraction-limited, submicron lateral resolution by switching to a CCD with smaller pixel size and/or increasing the transverse magnification. Moreover, measuring the LSPR wavelengths of PNPs is not limited to the substrate surface as in surface plasmon resonance imaging on thin films, which allows using PNPs as sensors to observe refractive index changes on the top surface of cell membrane or inside cells [11,12,20]. Therefore, the HSIS could potentially be applied to high-resolution imaging of PNPs in living cells to obtain information about localized molecular interactions.
The reported HSIS exploits sensitive yet economic large CCD chips to capture wide-field images, thanks to mass production of such devices. A hyperspectral image of immobilized PNPs can be obtained in 5 seconds with a field of view of 659 μm × 496 μm and spectral resolution of 21.4 nm at 532 nm. We proposed a simple method to empirically determine the OPD step size for each pixel within the field of view and achieved accurate wavelength calibration. We successfully demonstrated the application of the HSIS system for efficient detection of PNPs both as multicolor tags for microarrays and nanosensors for environmental refractive index. We believe that the integration of data analysis methods with high-throughput optical detection techniques such as the reported HSIS would facilitate more widespread use of PNPs for biomedical applications including high-spatial-resolution spectral imaging and high-throughput detection of biomolecules using microarrays.
Acknowledgements
The authors thank the National Science Council for financial support (grants 96-2627-B-002-013, 97-2627-B-002-009, and 98-2627-B-002-006) of this research, and Yi-Shen Lee and Chung-Jui Wu for help with sample preparation.
References and links
1. J. Reichert, A. Csáki, J. M. Köhler, and W. Fritzsche, “Chip-based optical detection of DNA hybridization by means of nanobead labeling,” Anal. Chem. 72(24), 6025–6029 (2000). [CrossRef]
2. T. A. Taton, G. Lu, and C. A. Mirkin, “Two-color labeling of oligonucleotide arrays via size-selective scattering of nanoparticle probes,” J. Am. Chem. Soc. 123(21), 5164–5165 (2001). [CrossRef] [PubMed]
3. S. J. Oldenburg, C. C. Genick, K. A. Clark, and D. A. Schultz, “Base pair mismatch recognition using plasmon resonant particle labels,” Anal. Biochem. 309(1), 109–116 (2002). [CrossRef] [PubMed]
4. S. Schultz, D. R. Smith, J. J. Mock, and D. A. Schultz, “Single-target molecule detection with nonbleaching multicolor optical immunolabels,” Proc. Natl. Acad. Sci. U.S.A. 97(3), 996–1001 (2000). [CrossRef] [PubMed]
5. K. Sokolov, M. Follen, J. Aaron, I. Pavlova, A. Malpica, R. Lotan, and R. Richards-Kortum, “Real-time vital optical imaging of precancer using anti-epidermal growth factor receptor antibodies conjugated to gold nanoparticles,” Cancer Res. 63(9), 1999–2004 (2003). [PubMed]
6. I. H. El-Sayed, X. Huang, and M. A. El-Sayed, “Surface plasmon resonance scattering and absorption of anti-EGFR antibody conjugated gold nanoparticles in cancer diagnostics: applications in oral cancer,” Nano Lett. 5(5), 829–834 (2005). [CrossRef] [PubMed]
7. J. Yguerabide and E. E. Yguerabide, “Light-scattering submicroscopic particles as highly fluorescent analogs and their use as tracer labels in clinical and biological applications,” Anal. Biochem. 262(2), 157–176 (1998). [CrossRef] [PubMed]
8. C. Sönnichsen, B. M. Reinhard, J. Liphardt, and A. P. Alivisatos, “A molecular ruler based on plasmon coupling of single gold and silver nanoparticles,” Nat. Biotechnol. 23(6), 741–745 (2005). [CrossRef] [PubMed]
9. J. Aaron, N. Nitin, K. Travis, S. Kumar, T. Collier, S. Y. Park, M. José-Yacamán, L. Coghlan, M. Follen, R. Richards-Kortum, and K. Sokolov, “Plasmon resonance coupling of metal nanoparticles for molecular imaging of carcinogenesis in vivo,” J. Biomed. Opt. 12(3), 034007 (2007). [CrossRef] [PubMed]
10. J. N. Anker, W. P. Hall, O. Lyandres, N. C. Shah, J. Zhao, and R. P. Van Duyne, “Biosensing with plasmonic nanosensors,” Nat. Mater. 7(6), 442–453 (2008). [CrossRef] [PubMed]
11. A. Wax and K. Sokolov, “Molecular imaging and darkfield microspectroscopy of live cells using gold plasmonic nanoparticles,” Laser Photon. Rev. 3(1-2), 146–158 (2009). [CrossRef]
12. M. J. Crow, G. Grant, J. M. Provenzale, and A. Wax, “Molecular imaging and quantitative measurement of epidermal growth factor receptor expression in live cancer cells using immunolabeled gold nanoparticles,” AJR Am. J. Roentgenol. 192(4), 1021–1028 (2009). [CrossRef] [PubMed]
13. J. Becker, O. Schubert, and C. Sönnichsen, “Gold nanoparticle growth monitored in situ using a novel fast optical single-particle spectroscopy method,” Nano Lett. 7(6), 1664–1669 (2007). [CrossRef] [PubMed]
14. G. L. Liu, J. C. Doll, and L. P. Lee, “High-speed multispectral imaging of nanoplasmonic array,” Opt. Express 13(21), 8520–8525 (2005). [CrossRef] [PubMed]
15. A. Curry, G. Nusz, A. Chilkoti, and A. Wax, “Substrate effect on refractive index dependence of plasmon resonance for individual silver nanoparticles observed using darkfield microspectroscopy,” Opt. Express 13(7), 2668–2677 (2005). [CrossRef] [PubMed]
16. K. L. Kelly, E. Coronado, L. L. Zhao, and G. C. Schatz, “The optical properties of metal nanoparticles: The influence of size, shape, and dielectric environment,” J. Phys. Chem. B 107(3), 668–677 (2003). [CrossRef]
17. D. M. Haaland, J. A. Timlin, M. B. Sinclair, M. H. V. Benthem, M. J. Martinez, A. D. Aragon, and M. Werner-Washburne, “Multivariate curve resolution for hyperspectral image analysis: applications to microarray technology,” in Proc. SPIE, R. M. Levenson, G. H. Bearman, and A. Mahadevan-Jansen, eds. (San Jose, CA, USA 2003), p. 55.
18. J. A. Timlin, D. M. Haaland, M. B. Sinclair, A. D. Aragon, M. J. Martinez, and M. Werner-Washburne, “Hyperspectral microarray scanning: impact on the accuracy and reliability of gene expression data,” BMC Genomics 6(1), 72 (2005). [CrossRef] [PubMed]
19. A. Dubois, J. Moreau, and C. Boccara, “Spectroscopic ultrahigh-resolution full-field optical coherence microscopy,” Opt. Express 16(21), 17082–17091 (2008). [CrossRef] [PubMed]
20. A. C. Curry, M. Crow, and A. Wax, “Molecular imaging of epidermal growth factor receptor in live cells with refractive index sensitivity using dark-field microspectroscopy and immunotargeted nanoparticles,” J. Biomed. Opt. 13(1), 014022 (2008). [CrossRef] [PubMed]
21. B. Ford, M. Descour, and R. Lynch, “Large-image-format computed tomography imaging spectrometer for fluorescence microscopy,” Opt. Express 9(9), 444–453 (2001). [CrossRef] [PubMed]
22. R. T. Kester, L. Gao, and T. S. Tkaczyk, “Development of image mappers for hyperspectral biomedical imaging applications,” Appl. Opt. 49(10), 1886–1899 (2010). [CrossRef] [PubMed]
23. G. Steiner, “Surface plasmon resonance imaging,” Anal. Bioanal. Chem. 379(3), 328–331 (2004). [CrossRef] [PubMed]
24. R. Vander and S. G. Lipson, “High-resolution surface-plasmon resonance real-time imaging,” Opt. Lett. 34(1), 37–39 (2009). [CrossRef]
