  
|
||||||||||||
|
||||||||||||
Spectral Domain OCT and Image Scattering Contrast Agents OCT and Fluorescence Imaging of VEGF/Cy5.5 in a Mouse Model of Colorectal Cancer Vascular growth, or angiogenesis, is considered an early-stage warning sign of most malignant cancers. Dr. Joseph Baker and team have developed a molecule, single chain (sc) VEGF, conjugated to a fluorescent dye, Cy5.5, to detect VEGF receptor sites via fluorescence imaging. Our endoscopic, minimally invasive imaging systems combine optical coherence tomography (OCT) and laser-induced fluorescence (LIF) to obtain structural and biochemical information from tissue in vivo. In previous studies of colorectal cancer detection in mice, we have achieved 95% sensitivity for early-stage tumors in the colon using the structural data provided by OCT. By using OCT to identify tumors, fluorescence, or LIF, data from VEGF/Cy5.5 can be correlated with tumor location. We find that some tumors apparently fluoresce more than others. While these tumors are benign, there is currently no method to identify those tumors that will remain benign and those that will continue to grow and become malignant. Variability in VEGF expression suggests that LIF imaging of VEGF/Cy5.5 may provide information about a tumor’s potential malignancy. Unfortunately, many sites outside the colon express fluorescence due to VEGF/Cy5.5, including small intestine and fat. These confounders necessitate appropriate engineering and VEGF/Cy5.5 administration to improve specificity for tumor sites. Most tumors are epithelial, whereas, most confounders are external to the colon. Topically applying VEGF/Cy5.5 significantly improves specificity for tumors as compared with intravenous injection, without sacrificing sensitivity. Furthermore, increasing excitation power at the surface may improve specificity for tumors. Possible improvements to the LIF imaging system may include using a high NA focusing optic or using a shorter wavelength to limit depth of penetration. While this study focuses on the application of VEGF/Cy5.5 enhanced LIF imaging in mice, fluorescence imaging with VEGF/Cy5.5 can easily be incorporated in human colonoscopies. Autofluorescence imaging in the human colon has been studied as early as the 1990s with high sensitivity and specificity for dysplasia and rectal cancer. By using an exogenous fluorophore, Cy5.5, conjugated to VEGF, pre-cancers may be identified in addition to more severe disease, potentially improving colorectal cancer prevention.
Gold Nanoshell Enhancement of Tumors in OCT Imaging Gold nanoshells are nanoparticles that consist of a dielectric core (silica) and gold outer layer. Due to their size, shape, and material properties, they exhibit a localized plasmon resonance, which can be spectrally tuned by changing their geometry. More specifically, the inner dielectric core diameter to total diameter affects the resonant wavelength. When this resonant wavelength is tuned to the near-infrared (NIR), the nanoshells scatter and absorb in the NIR more efficiently than tissue, enabling contrast enhancement (due to increased scattering) and photothermal therapy (due to increased absorption). Contrast due to increased scattering and absorption can be observed with optical coherence tomography (OCT). In a region of nanoshells, both effects cause a heightened attenuation with depth, as more light is either scattered or absorbed and less light is available to penetrate deeper. In this study, nanoshells are conjugated to epidermal growth factor receptor (EGFR) to make them tumor-specific. These nanoshells are topically applied in a mouse colon, which has been treated with a carcinogen, AOM, to induce colorectal tumors. We show heightened attenuation in OCT images in regions of abnormal growth in vivo. Furthermore, these regions can be identified quantitatively by examining the slope, surface brightness to average brightness ratio, and regression analysis.
Spectral domain optical coherence tomography (SD-OCT) Traditional (time domain) OCT is accomplished by low coherence interferometry, in which a signal is only collected when the two arms of the interferometer are matched. To collect depth scans (or ascans), a reference mirror in one of the interferometer arms is dithered. This motion causes the coherence gate to dither in depth in a sample in the other interferometer arm.
The speed of time domain OCT systems is typically limited by the speed of the reference mirror, on the order of tens of ascans per second. SD-OCT can far surpass this rate, to thousands of ascans per second and higher, by eliminating the need to scan a reference mirror. To collect depth scans, the interferometer signal is collected as a function of wavelength. The following is the formula for interference of two waves for a single wavelength. ‘I’ represents intensity and ‘opd’ optical path difference.
If inverse wavelength, or wavenumber, is considered to be the variable, which is convenient since the data is collected as a function of wavelength, note that each optical path difference (opd) results in a different frequency. In OCT, the opd is equal to twice the distance from the zero delay, or coherence gate location, to the sample depth location of interest. Therefore, a depth scan can be visualized by Fourier Transforming the spectral data as a function of wavenumber.
One method of collecting spectral data is to replace the photodiode detector in time domain OCT with a spectrometer. One of the convenient aspects of this layout is that it easily facilitates switching between time domain and spectral domain OCT by simply moving the detection fiber output from the photodiode detector, used for time domain OCT, to the spectrometer input, used for SD-OCT. Here are some of the main specifications of the system in our laboratory.
Specifications:
Spectrometer Layout
Custom Focusing Lens Layout
Lens MTF at 900 nm
Interesting Effects in SD-OCT Roll-off Unlike in time domain OCT, SD-OCT images experience an inherent decrease in signal with depth. This roll-off with depth has nothing to do with the sample being imaged but with the detection hardware itself. To use the example of our system, in which the detector is a spectrometer, the CCD array acts as a low pass filter by summing the signal over a certain range of wavelengths, so that high frequencies, which correspond to deeper layers in the sample, are gradually filtered out. This effect is visualized in the MTF curve. The MTF, or Modulation Transfer Function, visually demonstrates image degradation with increasing frequency. Our cut-off is about 35.7 inverse millimeters, which corresponds to the Nyquist frequency from finite sampling as well as signal degradation from pixel summing of the CCD array. The pixels are spaced 14 mm apart so the Nyquist frequency, or highest recreatable frequency, is 1/(2*14 mm) = 35.7 mm -1. For a optical bandwidth of 200 nm, wavelengths 800-1000 nm, dispersed on this array of 2048 pixels, the spatial frequency, 35.7 mm -1, corresponds to a period in wavelength of 200 nm/2048 * 2 = 0.2 nm. Converting this period in wavelength to a period in wavenumber at 900 nm gives 0.2 nm/(900*900 nm 2) = 2.47e-7 nm -1. This period in wavenumber corresponds to a frequency in wavenumber of 1/(2.47e-7) nm = 4.05e6 nm = 4.05 mm. Therefore, the maximum detectable opd is about 4 mm, which corresponds to a maximum imaging depth of about 2 mm. Roll-off can also be thought of in terms of coherence length. Each pixel on the CCD array responds to a certain range of wavelengths. For our system, the total bandwidth, 200 nm, is dispersed over 2048 pixels so our wavelength range per pixel is approximately 200 nm / 2048 = 0.1 nm. The pixel coherence length, if we assume a center wavelength of 900 nm, is about 2*ln(2)/pi * 900 nm * 900 nm / 0.1 nm = 4 mm, which corresponds to 2 mm of imaging depth. The CCD array is not the only element that causes roll-off. Other elements in the system, the grating and focusing optic in particular, can contribute. Aberrations in the focusing optic and diffraction effects from the focusing optic and grating degrade the wavelength resolution, which results in roll-off.
Calibration and Sample Dispersion One convenient feature of SD-OCT is that we have direct access to phase information as a function of wavelength. Having the data in this format allows relatively straightforward dispersion correction. To correct system dispersion, many groups have developed iterative methods to calculate a phase correction term as a function of wavelength. The procedure necessitates imaging the spectrum with a single scattering in the sample arm, such that only one frequency should be present across the spectrum. To obtain phase information, the Hilbert Transform is used to calculate the imaginary part of the signal. Since only one frequency is present across the signal, the phase is the inverse tangent of the imaginary part, produced by the Hilbert Transform, divided by the real part, the actual signal on the spectrometer. The phase of the signal should be 2*pi/wavelength *opd, which is linear with respect to inverse wavelength, or wavenumber. In the absence of system dispersion, the phase plot will only be linear if the spectrum samples are measured in even increments of wavenumber. Unless the spectrometer has been designed to disperse evenly in wavenumber, this will not be the case. Taking the Fourier Transform at this point will produce erroneous results because a nonlinear phase corresponds to multiple frequencies, despite the fact that only one scatterer is present. To correct this, resampling is required to produce samples evenly spaced in wavenumber, which necessarily is the case when resampling produces a linear phase. Interestingly, if the single scatterer used in this experiment is immersed in a dispersive media, where the dispersive media extends from the zero delay, or coherence gate, downward to the scatterer position, the resultant resampling will be different than if the scatterer is in air. This effect occurs because media dispersion is accounted for by resampling. The idea behind media dispersion correction is that the wavelength of light changes when entering a material with index of refraction not equal to one. So a scatterer that used to give 1 wave of path difference at 900 nm is now 1 wave of path difference at the wavelength which when divided by the index gives 900 nm. The previous algorithm results in resampling as if this new wavelength is 900 nm, as opposed to its actual wavelength in a vacuum. In other words, the index of refraction per wavelength is accounted for in this algorithm, so media dispersion is corrected inherently *. * A. R. Tumlinson, B. Hofer, A. M. Winkler, B. Považay, W. Drexler, J. K. Barton. “Inherent Media Dispersion Compensation in Frequency Domain Optical Coherence Tomography by Accurate k-sampling.” Applied Optics 47 (5), pp. 687-693 (Feb. 2008).
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Email Dr. Barton Email the Webmaster |
Phone: 520.621.4116 Fax: 520.626.4824 |
1657 E Helen Street #124 Tucson, AZ 85719 |
