Hyperspectral band selection is the process of selecting an optimal set of narrow wavelength bands from a large number over a broad range, typically for one of two purposes: hyperspectral reconstruction or classification. The former seeks to condense the information content of the full resolution spectrum so that the spectrum may be reconstructed from a relatively small subset of wavelength bands. The latter seeks to enable classification based on features contained within this small subset. In this paper, we introduce a new approach for automated band selection based on analysis of the weight planes from a trained self-organizing map. We refer to this approach as the self-organizing map weight plane distance (SOM WPD) method. We evaluate its benefits by using it to select optimal visible/near infrared (VNIR) and fluorescence wavelength bands from a recent fish fraud study where hyperspectral imaging was used to identify the true species of fish fillets. We apply four common machine learning classifiers to perform this species classification and compare the results to those obtained using a genetic algorithm-based method. This latter method optimized band selection for hyperspectral reconstruction, and these same bands were in turn used for classification. The findings presented in this paper show great promise for the SOM WPD which produced higher classification accuracy with two of the four classifiers for the VNIR data and with all four classifiers for the fluorescence data as compared with the genetic algorithm-based method.
We introduce a multimode dermoscope (SkinSpectTM) we developed for early detection of
melanoma by combining fluorescence, polarization and hyperspectral imaging. Acquired reflection
image datacubes were input to a wavelength-dependent linear model to extract the relative
contributions of skin chromophores at every pixel. The oxy-hemoglobin, deoxy hemoglobin,
melanin concentrations, and hemoglobin oxygen saturation by the single step linear least square
fitting and Kubelka-Munk tissue model using cross polarization data cubes were presented. The
comprehensive data obtained by SkinSpect can be utilized to improve the accuracy of skin
chromophore decomposition algorithm with less computation cost. As an example in this work, the
deoxy-hemoglobin over-estimation error in highly pigmented lesion due to melanin and deoxy
hemoglobin spectral cross talk were analyzed and corrected using two-step linear least square fitting
procedure at different wavelength ranges. The proposed method also tested in skin with underlying
vein area for validating the proof of concept.
We present an approach for rapidly and quantitatively mapping tissue absorption and scattering spectra in a wide-field, noncontact imaging geometry by combining multifrequency spatial frequency domain imaging (SFDI) with a computed-tomography imaging spectrometer (CTIS). SFDI overcomes the need to spatially scan a source, and is based on the projection and analysis of periodic structured illumination patterns. CTIS provides a throughput advantage by simultaneously diffracting multiple spectral images onto a single CCD chip to gather spectra at every pixel of the image, thus providing spatial and spectral information in a single snapshot. The spatial-spectral data set was acquired 30 times faster than with our wavelength-scanning liquid crystal tunable filter camera, even though it is not yet optimized for speed. Here we demonstrate that the combined SFDI-CTIS is capable of rapid, multispectral imaging of tissue absorption and scattering in a noncontact, nonscanning platform. The combined system was validated for 36 wavelengths between 650-1000 nm in tissue simulating phantoms over a range of tissue-like absorption and scattering properties. The average percent error for the range of absorption coefficients (μa) was less than 10% from 650-800 nm, and less than 20% from 800-1000 nm. The average percent error in reduced scattering coefficients (μs′) was less than 5% from 650-700 nm and less than 3% from 700-1000 nm. The SFDI-CTIS platform was applied to a mouse model of brain injury in order to demonstrate the utility of this approach in characterizing spatially and spectrally varying tissue optical properties.
Retinal imaging spectroscopy can provide functional maps using chromophore spectra. For example, oxygen saturation maps show ischemic areas from diabetes and venous occlusions. Obtaining retinal spatial-spectral data has been difficult due to saccades and long data acquisition times (>5 s). We present a snapshot imaging spectrometer with far-reaching applicability that acquires a complete spatial-spectral image cube in ~3 ms from 450 to 700 nm with 50 bands, eliminating motion artifacts and pixel misregistration. Current retinal spectral imaging approaches are incapable of true snapshot operation over a wide spectral range with a large number of spectral bands. Coupled to a fundus camera, the instrument returns true color retinal images for comparison to standard fundus images and for image validation while the patient is still dilated. Oxygen saturation maps were obtained with a three-wavelength algorithm: for healthy subjects arteries were ~95% and veins 30 to 35% less. The instrument is now undergoing clinical trials.
We report on a volume holographic imaging spectrometer (VHIS) system which allows retrieval of a scene's two-dimensional spatial information as well as its spectral information. This is performed using a transmission volume hologram and a simple rotary scanning mechanism. The system has the advantages of high spectral and spatial resolutions and the potential of single-shot, four-dimensional (3D spatial plus 1D spectral) imaging by recording multiple volume holograms in the same material. Also, due to the transmission diffraction geometry, the system automatically eliminates the stray excitation light from the captured signal. We give theoretical analysis of the performance and experimental demonstration using fluorescent CdSe/ZeS quantum dots. The measured quantum dots spectra agree well with the spectra obtained using a conventional spectrometer.
The computed tomographic imaging spectrometer (CTIS) is a passive non-scanning instrument which simultaneously records a scenes spectral content as well as its 2-D spatial. Simultaneously implies a time frame limited only by the frame rate and signal-to-noise of the imaging device. CTIS accomplishes this by feeding incident scene radiation through a computer generated hologram (CGH) in Fourier space. The resulting dispersion pattern is recorded on a conventional pixilated imager and is stored on a local computer for post processing using iterative reconstruction techniques. A virtual 3-D datacube is constructed with one dimension in terms of energy weights for each wavelength band. CTIS is ideal for observing rapidly varying targets and has found use in military, bio-medical and astronomical applications. For the first time we have built an entirely reflective design based on the popular Offner reflector using a computer generated hologram formed on a convex mirror surface. Furthermore, a micro electro-mechanical system (MEMS) has been uniquely incorporated as a dynamic field stop for smart scene selection. Both the MEMS and reflective design are discussed. The CTIS multiplexes spatial and spectral information, so the two quantities are interdependent and adjustments must be made to the design in order to allow adequate sampling for our given application. Optical aberrations arising from a tilted image plane are alleviated through design optimization.
Multi-color fluorescence microscopy has become a popular way to discriminate between multiple proteins, organelles or functions in a single cell or animal and can be used to approximate the physical relationships between individual proteins within the cell, for instance, by using Fluorescence Resonance Energy Transfer (FRET). However, as researchers attempt to gain more information from single samples by using multiple dyes or fluorescent proteins (FPs), spectral overlap between emission signals can obscure the data. Signal separation using glass filters is often impractical for many dye combinations. In cases where there is extensive overlap between fluorochromes, separation is often physically impossible or can only be achieved by sacrificing signal intensity. Here we test the performance of a new, integrated laser scanning system for multispectral imaging, the Zeiss LSM 510 META. This system consists of a sensitive multispectral imager and online linear unmixing functions integrated into the system software. Below we describe the design of the META device and show results from tests of the linear unmixing experiments using fluorochromes with overlapping emission spectra. These studies show that it is possible to expand the number of dyes used in multicolor applications.
Optical sensing of biomolecules on microfabricated glass surfaces requires surface coatings that minimize nonspecific binding while preserving the optical properties of the sensor. Microspheres with whispering-gallery (WG) modes can achieve quality factor (Q) levels many orders of magnitude greater than those of other WG-based microsensors: greater than 1010 in air, and greater than 109 in a variety of solvents, including methanol, H2O and phosphate buffered saline (PBS). The presence of dyes that absorb in the wavelength of the WG excitation in the evanescent zone can cause this Q value to drop by almost 3 orders of magnitude. Silanization of the surface with mercapto-terminal silanes is compatible with high Q (>109), but chemical cross-linking of streptavidin reduces the Q to 105-106 due to build-up of a thick, irregular layer of protein. However, linkage of biotin to the silane terminus preserves the Q at a ~2x107 and yields a reactive surface sensitive to avidin-containing ligands in a concentration-dependent manner. Improvements in the reliability of the surface chemistry show promise for construction of an ultrasensitive biosensor.
When imaging the backscattered light from turbid tissue using a broadband illumination source, the random scattering of photons within the tissue causes wavelength-dependent optical coupling between pixels. That is, a photon may exit the tissue surface an extended distance away from its entry point. The resulting spectral crosstalk in the detected image can be explained by studying the mean photon path lengths through the tissue. Considering complex tissue geometries with features such as cylindrical vessels, these photons not only travel multiple paths due to wavelength- dependent absorption and scattering, but may also travel through multiple chromophores. To study the effects of 3D features in object space on backscattered light into the image plane, we have constructed a Monte Carlo simulation capable of modeling 3D photon propagation for a tissue slab with an embedded cylinder. The results of hemoglobin-bearing vessels as a primary chromophore are investigated. Because of the relationship between mean photon path length and photon exit angle, we have shown that the choice of entrance pupil in the imaging system plays an important role on the detected backscatter for the specific case of embedded cylinders.
Rupture of atherosclerotic plaques - the main cause of heart attach and stokes - is not predictable. Hence even treadmill stress tests fail to detect many persons at risk. Fatal plaques are found at autopsies to be associated with active inflammatory cells. Classically, inflammation is detected by its swelling, red color, pain and heat. We have found that heat accurately locates the dangerous plaques that are significantly warmer then atherosclerotic plaques without the same inflammation. In order to develop a non-surgical method of locating these plaques, an IR fiber optic imaging system has been developed in our laboratory to evalute the causes and effect of heat in atherosclerotic plaques. The fiber optical imagin bundle consists of 900 individual As2S3 chalcogenide glass fibers which transmit IR radiation from 0.7 micrometers 7 micrometers with little energy loss. By combining that with a highly sensitive Indium Antimonide IR focal plane array detector, we are able to obtain thermal graphic images in situ. The temperature heterogeneity of atherosclerotic plaques developed in the arteral of the experimental animal models is under study with the new device. The preliminary experimental results from the animal model are encouraging. The potential of using this new technology in diagnostic evaluation of the vulnerable atherosclerotic plaques is considerable.
Spectral imaging permits two-dimensional mapping of the reflectance properties of biological systems. However, imaging in turbid media involves pixel sizes that are comparable to or smaller than the mean photon path length. This implies that the spectrum measured at a given pixel in the image plane will be determined by manifold photon trajectories through an extended volume in the object, so there is not a uniquely defined path length. In addition, this implies nonlinear spectral mixing for systems with multiple layers and chromophores. Using Monte Carlo model, we have studied photon path distributions in the case of layered turbid systems and their effects on spectral imaging. In particular, we emphasize the effect of hemoglobin on imaging reflectance-mode hyperspectral data.
We describe fluorescence spectral-imaging results with the computed-tomography imaging spectrometer (CTIS). This imaging spectrometer is capable of recording spatial and spectral data simultaneously. Consequently, the CTIS can be used to image dynamic phenomena involving multiple, spectrally overlapping fluorescence probes. This system is also optimal for simultaneously monitoring changes in spectral characteristics of multiple probes from different locations within the same sample. This advantage will provide additional information about the physiological changes in function form populations of cells which respond in a heterogeneous manner. The results presented in this paper consist of proof-of-concept imaging results from the CTIS in combination with two different systems of fore- optics. In the first configuration, raw image data were collected using the CTIS coupled to an inverted fluorescence microscope. The second configuration combined the CTIS with a confocal microscope equipped with a fiber-optic imaging bundle, previously for in vivo imaging. Image data were collected at frame rates of 15 frame per second and emission spectra were sample at 10-nm intervals with a minimum of 29 spectral bands. The smallest spatial sampling interval presented in this paper is 0.7 micrometers .
Radiometric models have been used to optimize instrument design or evaluate impacts of changes to the design during integration and test. Tradeoffs such as spectral and spatial resolution, telescope and spectrometer temperature, aperture, f/No., integration time, optics and filter transmissions, and so forth can be quickly changed to evaluate changes to the signal/noise ratio or other performance metrics. An alternative use of such models is to identify promising instrument proposals for further study. A series of models were constructed to evaluate general instrument designs as an illustration of this process. These models included two grating spectrometers and a spatially modulated interferometer. All were given a common set of radiometric inputs and telescope optical prescription. Result of the modeling illustrate the performance differences between instrument types, although signal/noise predictions should be evaluated along with other parameters such as manufacturability, precision of calibration, and so forth. Such modeling allows instrument developers to demonstrate to potential customers improvements in their instruments, and the advantages of their product over other instruments for a specific application. If a common set of inputs is used for the different instrument models, this technique gives customers one metric with which to evaluate the disparate proposals.
We describe fluorescence spectral-imaging results with the microscope computed-tomography imaging spectrometer ((mu) CTIS). This imaging spectrometer is capable of recording spatial and spectral data simultaneously. Consequently, the (mu) CTIS can be used to image dynamic phenomena involving multiple, spectrally overlapping fluorescence probes. The result presented in this paper consists of proof-of-concept imaging result using two static targets. The first is composed of 6-micrometers fluorescing microspheres and the second consists of rat sinusoid epithelial cells loaded with 0.5-micrometers fluorescing microspheres. Image data were collected in integration times of 16 msec, comparable to video frame rate integration times. The emission spectra were sampled at 10-nm intervals between 420 nm and 710 nm. The smallest spatial sampling interval presented in this paper is 1.7 micrometers .
We describe fluorescence spectral-imaging results with the microscope computed-tomography imaging spectrometer ((mu) CTIS). This imaging spectrometer is capable of recording spatial and spectral data simultaneously. Consequently, (mu) CTIS can be used to image dynamic phenomena. The results presented in this paper consist of imaging results using static targets consisting of 1 micrometers and 6 micrometers fluorescing microspheres. The emission spectra were sampled at a 10-nm interval between 430 nm and 710 nm. The smallest spatial sampling interval presented in this paper is 1.7 micrometers . Image data were collected in integration times of 16 msec.
Imaging spectrometers have recently moved out of the spaceflight environment, in which they were developed, to a host of other applications. Some of these new uses include the graphics and printing industry, process control, bio-medicine, clinical diagnostics and agriculture. For any of these applications, new approaches are necessary to design compact, portable instruments that can be easily and reliably calibrated. This paper presents one such implementation of an imaging spectrometer suitable for field use.
InSb photodiodes were examined for performance degradation after pulsed laser illumination at 0.532 micron and 1.064 micron. Incident laser powers ranged from 6 x 10 exp-18 micron-watts to 16 micron-watts in a 50 pm diameter spot. Dark current and spectral response were both measured before and after illumination. Dark current measurements were taken with the diode blanked off and viewing only 77 K surfaces. Long term stability tests demonstrated that the blackbody did not exhibit long term drifts. Other tests showed that room temperature variations did not affect the diode signal chain or the digitization electronics used in data acquisition. Results of the experiment show that the diodes did not exhibit changes in dark current or spectral response performance as a result of the laser illumination. A typical change in diode spectral response (before/after laser exposure) was about 0.2 percent +/- 0.2 percent.