The formation of bacterial colonies and biofilms requires coordinated gene expression, regulated cell differentiation,
autoaggregation, and intercellular communication. Therefore colonies of bacteria have been recognized as multicellular
organisms or "superorganisms." It has consequently been postulated that the phenotype of colonies formed by
microorganisms can be automatically recognized and classified using optical systems capable of collecting information
related to cellular pattern formation and morphology of colonies. Recently we have reported a first practical
implementation of such a system, capable of noninvasive, label-free classification and recognition of pathogenic Listeria
species. The design employed computer-vision and pattern-recognition techniques to classify scatter patterns produced
by bacterial colonies irradiated with laser light. Herein we report our efforts to extend this system to other genera of
bacteria such as Salmonella, Vibrio, Staphylococcus, and E. coli. Application of orthogonal moments, as well as texture
descriptors for image feature extraction, provides high robustness in the presence of noise. An improved pattern
classification scheme based on an SVM algorithm provides better results than the previously employed neural network
system. Low error rates determined by cross-validation, reproducibility of the measurements, and overall robustness of
the recognition system prove that the proposed technology can be implemented in automated devices for bacterial
detection.
Bacterial contamination of food products puts the public at risk and also generates a substantial cost for the food-processing industry. One of the greatest challenges in the response to these incidents is rapid recognition of the bacterial agents involved. Only a few currently available technologies allow testing to be performed outside of specialized microbiological laboratories. Most current systems are based on the use of expensive PCR or antibody-based techniques, and require complicated sample preparation for reliable results. Herein, we report our efforts to develop a noninvasive optical forward-scattering system for rapid, automated identification of bacterial colonies grown on solid surfaces. The presented system employs computer-vision and pattern-recognition techniques to classify scatter patterns produced by bacterial colonies irradiated with laser light. Application of Zernike and Chebyshev moments, as well as Haralick texture descriptors for image feature extraction, allows for a very high recognition rate. An SVM algorithm was used for classification of patterns. Low error rates determined by cross-validation, reproducibility of the measurements, and robustness of the system prove that the proposed technology can be implemented in automated devices for bacterial detection.
Shear stress is known to have a significant effect on the state of cellular differentiation. It also induces morphologic responses including changes to cytoskeletal organization subsequently leading to changes in cell shape. In fact, fluid shear stress caused by blood flow is a major determinant of vascular remodeling and can lead to development of atherosclerosis. The morphological changes are usually evaluated using boundary-based shape descriptors or binary geometrical moments on manually segmented cells. Although any one of the many automated segmentation methods could be employed, these techniques are known to be complex and time consuming, and often require user input to operate properly, which is especially problematic for HCS systems. Therefore, development of robust, quantitative morphological measurements that are not dependent on precision and reproducibility of segmentation is extremely important for a substantial improvement of shear-stress analysis. The goals of this study were to find simple morphological descriptors that could be applied to cells isolated by tessellation in order to enable a high-throughput screening of morphological shear-stress response, and to determine the amount of fluid shear stress to which endothelial cells were exposed on the basis of changes in their morphology. The proposed technique is based on the monitoring of changes in cytoskeleton organization using texture descriptors, rather than on quantifying cell-boundary modifications. We showed that objects identified by Voronoi tessellation carried enough information about cytoskeleton texture of individual cells to create a robust classifier. Our approach provided higher discriminant and predictive powers, and better classification capability, than traditional boundary-based methods. The robustness of classification in the presence of segmentation difficulties makes the proposed approach particularly suitable for automated HCS systems.
Traditional biological and chemical methods for pathogen identification require complicated sample preparation for
reliable results. Optical scattering technology has been used for identification of bacterial cells in suspension, but with
only limited success. Our published reports have demonstrated that scattered light based identification of Listeria
colonies growing on solid surfaces is feasible with proper pattern recognition tools. Recently we have extended this
technique to classification of other bacterial genera including, Salmonella, Bacillus, and Vibrio. Our approach may be highly applicable to early detection and classification of pathogens in food-processing industry and in healthcare.
The unique scattering patterns formed by colonies of different species are created through differences in colony
microstructure (on the order of wavelength used), bulk optical properties, and the macroscopic morphology. While it is
difficult to model the effect on scatter-signal patterns owing to the microstructural changes, the influence of bulk optical
properties and overall shape of colonies can be modeled using geometrical optics. Our latest research shows that it is
possible to model the scatter pattern of bacterial colonies using solid-element optical modeling software (TracePro), and
theoretically assess changes in macro structure and bulk refractive indices. This study allows predicting the theoretical
limits of resolution and sensitivity of our detection and classification methods. Moreover, quantification of changes in
macro morphology and bulk refractive index provides an opportunity to study the response of colonies to various
reagents and antibiotics.
Bacterial contamination by Listeria monocytogenes puts the public at risk and is also costly for the food-processing
industry. Traditional methods for pathogen identification require complicated sample preparation for reliable results.
Previously, we have reported development of a noninvasive optical forward-scattering system for rapid identification of
Listeria colonies grown on solid surfaces. The presented system included application of computer-vision and patternrecognition
techniques to classify scatter pattern formed by bacterial colonies irradiated with laser light. This report
shows an extension of the proposed method. A new scatterometer equipped with a high-resolution CCD chip and
application of two additional sets of image features for classification allow for higher accuracy and lower error rates.
Features based on Zernike moments are supplemented by Tchebichef moments, and Haralick texture descriptors in the
new version of the algorithm. Fisher's criterion has been used for feature selection to decrease the training time of
machine learning systems. An algorithm based on support vector machines was used for classification of patterns. Low
error rates determined by cross-validation, reproducibility of the measurements, and robustness of the system prove that
the proposed technology can be implemented in automated devices for detection and classification of pathogenic bacteria.
Bacterial contamination by Listeria monocytogenes not only puts the public at risk, but also is costly for the food-processing industry. Traditional biochemical methods for pathogen identification require complicated sample preparation for reliable results. Optical scattering technology has been used for identification of bacterial cells in suspension, but with only limited success. Therefore, to improve the efficacy of the identification process using our novel imaging approach, we analyze bacterial colonies grown on solid surfaces. The work presented here demonstrates an application of computer-vision and pattern-recognition techniques to classify scatter patterns formed by Listeria colonies. Bacterial colonies are analyzed with a laser scatterometer. Features of circular scatter patterns formed by bacterial colonies illuminated by laser light are characterized using Zernike moment invariants. Principal component analysis and hierarchical clustering are performed on the results of feature extraction. Classification using linear discriminant analysis, partial least squares, and neural networks is capable of separating different strains of Listeria with a low error rate. The demonstrated system is also able to determine automatically the pathogenicity of bacteria on the basis of colony scatter patterns. We conclude that the obtained results are encouraging, and strongly suggest the feasibility of image-based biodetection systems.
Segmentation of dynamic PET images is an important preprocessing step for kinetic parameter estimation. A single time activity curve (TAC) is extracted for each segmented region. This TAC is then used to estimate the kinetic parameters of the segmented region. Current methods perform this task in two independent steps; first dynamic positron emission tomography (PET) images are reconstructed from the
projection data using conventional tomographic reconstruction methods, then the time activity curves (TAC) of the pixels are clustered into a predetermined number of clusters. In this paper, we propose to cluster the regions of dynamic PET images directly on the projection data and simultaneously estimate the TAC of each cluster.
This method does not require an intermediate step of tomographic reconstruction for each time frame. Therefore the dimensionality of the estimation problem is reduced. We compare the proposed method with weighted least squares (WLS) and expectation maximization with Gaussian mixtures methods (GMM-EM). Filtered backprojection is used to reconstruct the emission images required by these methods.
Our simulation results show that the proposed method can substantially decrease the number of mislabeled pixels and reduce the root mean squared error (RMSE) of the cluster TACs.
Discrete Chebyshev moments (due to discrete polynomial basis) do not have the discretization errors that continuous-domain Legendre and Zernike moments contain. Calculation of polynomial basis coefficients of discrete moments is generally performed using recurrence relationships. Such recurrence equations cause numerical error accumulation especially for calculation of higher-order moments and for larger image sizes, causing significant degradation of image reconstruction from these moments. A method for better image reconstruction from high orders of discrete Chebyshev moments is demonstrated. This is accomplished by calculating Chebyshev polynomial coefficients directly from their definition formulas using arbitrary precision arithmetic and by forming lookup tables from these coefficients.
Pathogenic bacterial contamination in food products is costly to the public and to industry. Traditional methods for detection and identification of major food-borne pathogens such as Listeria monocytogenes typically take 3-7 days. Herein, the use of optical scattering for rapid detection, characterization, and identification of bacteria is proposed. Scatter patterns produced by the colonies are recognized without the need to use any specific model of light scattering on biological material. A classification system was developed to characterize and identify the scatter patterns obtained from colonies of various species of Listeria. The proposed classification algorithm is based on Zernike moment invariants (features) calculated from the scatter images. It has also been demonstrated that even a simplest approach to multivariate analysis utilizing principal component analysis paired with clustering or linear discriminant analysis can be successfully used to discriminate and classify feature vectors computed from the bacterial scatter patterns.
Detailed information on cellular and sub-cellular interactions can be extracted from large-scale data sets through the application of image processing and analysis techniques from computer vision and pattern recognition. An automated, high-speed method for analysis of cellular systems in 2D includes boundary analysis of the cells and may be extended to texture (content) analysis or further. The overall goal of such analysis is to reach conclusions as to the physiological state and behavior of the cells. In this paper, we focus on shape analysis of cells, as shape is an effective factor for quantification of the many apparent physiological changes. We explore shape analysis techniques, including geometric (regular), Zernike, and Krawtchouk moment invariants. We also report on our investigation of the effects of resolution changes (in imaging systems) on the descriptors of cell shape in terms of stability and consistence of these moment invariants. Our results show that Krawtchouk moment invariants are better cell shape descriptors compared to geometric moment invariants in low resolution images.
Multispectral imaging has been in use for over half a century. Owing to advances in digital photographic technology, multispectral imaging is now used in settings ranging from clinical medicine to industrial quality control. Our efforts focus on the use of multispectral imaging coupled with spectral deconvolution for measurement of endogenous tissue fluorophores and for animal tissue analysis by multispectral fluorescence, absorbance, and reflectance data.
Multispectral reflectance and fluorescence images may be useful in evaluation of pathology in histological samples. For example, current hematoxylin/eosin diagnosis limits spectral analysis to shades of red and blue/grey. It is possible to extract much more information using multispectral techniques. To collect this information, a series of filters or a device such as an acousto-optical tunable filter (AOTF) or liquid-crystal filter (LCF) can be used with a CCD camera, enabling collection of images at many more wavelengths than is possible with a simple filter wheel. In multispectral data processing the “unmixing” of reflectance or fluorescence data and analysis and the classification based upon these spectra is required for any classification. In addition to multispectral techniques, extraction of topological information may be possible by reflectance deconvolution or multiple-angle imaging, which could aid in accurate diagnosis of skin lesions or isolation of specific biological components in tissue. The goal of these studies is to develop spectral signatures that will provide us with specific and verifiable tissue structure/function information. In addition, relatively complex classification techniques must be developed so that the data are of use to the end user.
Noise removal faces a challenge: Keeping the image details. Resolving the dilemma of two purposes (smoothing and keeping image features in tact) working inadvertently of each other was an almost impossible task until anisotropic dif-fusion (AD) was formally introduced by Perona and Malik (PM). AD favors intra-region smoothing over inter-region in piecewise smooth images. Many authors regularized the original PM algorithm to overcome its drawbacks. We compared the performance of denoising using such 'fundamental' AD algorithms and one of the most powerful multiresolution tools available today, namely, wavelet shrinkage. The AD algorithms here are called 'fundamental' in the sense that the regularized versions center around the original PM algorithm with minor changes to the logic. The algorithms are tested with different noise types and levels. On top of the visual inspection, two mathematical metrics are used for performance comparison: Signal-to-noise ratio (SNR) and universal image quality index (UIQI). We conclude that some of the regu-larized versions of PM algorithm (AD) perform comparably with wavelet shrinkage denoising. This saves a lot of compu-tational power. With this conclusion, we applied the better-performing fundamental AD algorithms to a new imaging modality: Optical Coherence Tomography (OCT).
Anisotropic diffusion (AD), first introduced by Perona and Malik (PM), provides image enhancement a strong benefit as it favors intra-region over inter-region smoothing. Early updates on the original PM algorithm focused on the cures for its drawbacks. Later some authors provided their own versions of AD techniques with a wide variety of applications. We surveyed the pros and cons of many fundamental AD techniques. To put our purpose into perspective, we compared the performances of fundamental AD algorithms to simple traditional filters and to more sophisticated tools such as wavelets. Visual inspection and two mathematical criteria are used for performance comparison: Signal-to-noise ratio (SNR) and universal image quality index (UIQI). We believe that a good overview of its simplicity and power will show the rightful reason of so much interest in anisotropic diffusion since its introduction.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.