Software for quantification of labeled bacteria from digital microscope images by automated image analysis

Automated image analysis software, CellC, was developed and validated for quantification of bacterial cells from digital microscope images. CellC enables automated enumeration of bacterial cells, comparison of total count and specific count images [e.g., 4',6-diamino-2-phenylindole (DAPI) and fluorescence in situ hybridization (FISH) images], and provides quantitative estimates of cell morphology. The software includes an intuitive graphical user interface that enables easy usage as well as sequential analysis of multiple images without user intervention. Validation of enumeration reveals correlation to be better than 0.98 when total bacterial counts by CellC are compared with manual enumeration, with all validated image types. The software is freely available and modifiable: the executable files and MATLAB source codes can be obtained at www. cs. tut.fi/sgn/csb/cellc.     

Confocal stereology and image analysis: methods for estimating geometrical characteristics of cells and tissues from three-dimensional confocal images

A short review of confocal stereology and three-dimensional image analysis is presented, pointing out the achievements accomplished in this field by the Department of Biomathematics (Institute of Physiology, Prague). One of the methods of confocal stereology, the fakir method for surface area estimation, developed by this laboratory, is described. Methods for automatic measurement of geometrical characteristics of microscopical structures, based on 3-D image processing or surface triangulation, are discussed and compared with interactive stereological methods. Three-dimensional reconstruction programs and software implementation of stereological and digital methods as well as their practical applications are presented. The future trends are discussed.     

Automated interpretation of subcellular patterns in fluorescence microscope images for location proteomics.

Proteomics, the large scale identification and characterization of many or all proteins expressed in a given cell type, has become a major area of biological research. In addition to information on protein sequence, structure and expression levels, knowledge of a protein's subcellular location is essential to a complete understanding of its functions. Currently, subcellular location patterns are routinely determined by visual inspection of fluorescence microscope images. We review here research aimed at creating systems for automated, systematic determination of location. These employ numerical feature extraction from images, feature reduction to identify the most useful features, and various supervised learning (classification) and unsupervised learning (clustering) methods. These methods have been shown to perform significantly better than human interpretation of the same images. When coupled with technologies for tagging large numbers of proteins and high-throughput microscope systems, the computational methods reviewed here enable the new subfield of location proteomics. This subfield will make critical contributions in two related areas. First, it will provide structured, high-resolution information on location to enable Systems Biology efforts to simulate cell behavior from the gene level on up. Second, it will provide tools for Cytomics projects aimed at characterizing the behaviors of all cell types before, during, and after the onset of various diseases.     

Comparison of the Papanicolaou and Feulgen staining methods for DNA quantification by image analysis.

The possibility of using archival cytology material to study the evolution of neoplastic disease with regard to DNA content abnormalities was investigated. The accuracy of measuring the integrity optical density (OD) of nuclei that correlates to DNA amounts of those nuclei, on slides stained by the Papanicolaou method, was assessed and compared with a standard Feulgen method. Our data on rat liver nuclei peritoneal washings from patients with ovarian cystadenofibromas and ovarian cystadenocarcinomas suggested that analysis of cytological material using the Papanicolaou method is not reliable and that destaining the slides followed by Feulgen staining provides an optimal and reliable method of DNA quantification.     

Line FRAP with the confocal laser scanning microscope for diffusion measurements in small regions of 3-D samples

We present a truly quantitative fluorescence recovery after photobleaching (FRAP) model for use with the confocal laser scanning microscope based on the photobleaching of a long line segment. The line FRAP method is developed to complement the disk FRAP method reported before. Although being more subject to the influence of noise, the line FRAP model has the advantage of a smaller bleach region, thus allowing for faster and more localized measurements of the diffusion coefficient and mobile fraction. The line FRAP model is also very well suited to examine directly the influence of the bleaching power on the effective bleaching resolution. We present the outline of the mathematical derivation, leading to a final analytical expression to calculate the fluorescence recovery. We examine the influence of the confocal aperture and the bleaching power on the measured diffusion coefficient to find the optimal experimental conditions for the line FRAP method. This will be done for R-phycoerythrin and FITC-dextrans of various molecular weights. The ability of the line FRAP method to measure correctly absolute diffusion coefficients in three-dimensional samples will be evaluated as well. Finally we show the application of the method to the simultaneous measurement of free green fluorescent protein diffusion in the cytoplasm and nucleus of living A549 cells.     

Automated image analysis for array hybridization experiments.

MOTIVATION: Image analysis is a major part of data evaluation for array hybridization experiments in molecular biology. The program presented here is designed to analyze automatically images from hybridization experiments with various arrangements: different kinds of probes (oligonucleotides or complex probes), different supports (nylon filters or glass slides), different labeling of probes (radioactively or fluorescently). The program is currently applied to oligonucleotide fingerprinting projects and complex hybridizations. The only precondition for the use of the program is that the targets are arrayed in a grid, which can be approximately transformed to an orthogonal equidistant grid by a projective mapping. RESULTS: We demonstrate that our program can cope with the following problems: global distortion of the grid, missing of grid nodes, local deviation of the spot from its specified grid position. This is checked by different quality measures. The image analysis of oligonucleotide fingerprint experiments on an entire genetic library is used, in clustering procedures, to group related clones together. The results show that the program yields automatically generated high quality input data for follow up analysis such as clustering procedures. AVAILABILITY: The executable files will be available upon request for academics.     

Automated fluorescence image cytometry. DNA quantification and detection of chlamydial infections

Digitized fluorescence microscopy in conjunction with automated image segmentation is a promising approach for screening clinical specimens quickly and reliably. This paper describes the hardware and software of a prototype image-based cytometer that can identify fluorescent objects, discriminate true objects from artifacts and divide overlapping pairs of objects. The use of this image cytometer is discussed for: (1) the measurement of the DNA ploidy distribution of isolated mature rat liver nuclei labeled with 4',6-diamidine-2-phenylindole; (2) the comparison of the DNA ploidy distributions of the same samples measured by image cytometry (ICM) and flow cytometry (FCM); and (3) the quantification of chlamydial infection by double labeling cells with antichlamydiae antibody and Hoechst 33258 for nuclear DNA analysis. Ploidy distributions measured by the automated image cytometer compared favorably to those obtained by FCM. All pairs of overlapping nuclei were automatically detected by an additional computer algorithm, and those pairs that were clearly more than one nucleus by visual inspection were correctly divided. The irregular morphology of the chlamydiae-infected cells meant that 26% of them were not correctly identified in the fluorescein-stained images (as judged by manual inspection), but all cells were nevertheless detected correctly from the images of the Hoechst-stained samples. Automated fluorescence ICM yielded results similar to those obtained with FCM and had the additional benefit of maintaining cell and tissue architecture while preserving the opportunity for subsequent manual inspection of the specimen.     

Objective threshold selection procedure (OTS) for segmentation of scanning laser confocal microscope images

The determination of volumes and interface areas from confocal laser scanning microscopy (CLSM) images requires the identification of component objects by segmentation. An automated method for the determination of segmentation thresholds for CLSM imaging of biofilms was developed. The procedure, named objective threshold selection (OTS), is a three-dimensional development of the approach introduced by the popular robust automatic threshold selection (RATS) method. OTS is based on the statistical properties of local gray-values and gradients in the image. By characterizing the dependence between a volumetric feature and the intensity threshold used for image segmentation, the former can be determined with an arbitrary confidence level, with no need for user intervention. The identification of an objective segmentation procedure renders the possibility for the full automation of volume and interfacial area measurement. Images from two distinct biofilm systems, acquired using different experimental techniques and instrumental setups were segmented by OTS to determine biofilm volume and interfacial area. The reliability of measurements for each case was analyzed to identify optimal procedure for image acquisition. The automated OTS method was shown to reproduce values obtained manually by an experienced operator.     

A method for analysis of lipid vesicle domain structure from confocal image data

Quantitative characterization of the lateral structure of curved membranes based on fluorescence microscopy requires knowledge of the fluorophore distribution on the surface. We present an image analysis approach for extraction of the fluorophore distribution on a spherical lipid vesicle from confocal imaging stacks. The technique involves projection of volumetric image data onto a triangulated surface mesh representation of the membrane, correction of photoselection effects and global motion of the vesicle during image acquisition and segmentation of the surface into domains using histograms. The analysis allows for investigation of the morphology and size distribution of domains on the surface.     

A multi-model approach to simultaneous segmentation and classification of heterogeneous populations of cell nuclei in 3D confocal microscope images.

Automated segmentation and morphometry of fluorescently labeled cell nuclei in batches of 3D confocal stacks is essential for quantitative studies. Model-based segmentation algorithms are attractive due to their robustness. Previous methods incorporated a single nuclear model. This is a limitation for tissues containing multiple cell types with different nuclear features. Improved segmentation for such tissues requires algorithms that permit multiple models to be used simultaneously. This requires a tight integration of classification and segmentation algorithms. Two or more nuclear models are constructed semiautomatically from user-provided training examples. Starting with an initial over-segmentation produced by a gradient-weighted watershed algorithm, a hierarchical fragment merging tree rooted at each object is built. Linear discriminant analysis is used to classify each candidate using multiple object models. On the basis of the selected class, a Bayesian score is computed. Fragment merging decisions are made by comparing the score with that of other candidates, and the scores of constituent fragments of each candidate. The overall segmentation accuracy was 93.7% and classification accuracy was 93.5%, respectively, on a diverse collection of images drawn from five different regions of the rat brain. The multi-model method was found to achieve high accuracy on nuclear segmentation and classification by correctly resolving ambiguities in clustered regions containing heterogeneous cell populations.     

Three-dimensional component labeling of digital confocal microscope images enumerates replication centers in BrdUrd labeled fibroblasts.

An algorithm (3Dlabel-1) for labeling three dimensional binary data sets has been developed. Serial optical sections are acquired using a confocal laser scanning microscope. After filtering and thresholding operations, contiguous elements in the three dimensional data sets are identified and labeled. The results are used to calculate the number and size of objects. Additionally, the labeled data are displayed by applying an algorithm (3Ddisp-1) to generate stereo pairs, in which label numbers are color coded and depth is cued by intensity. These procedures have been applied in investigations of the temporal and spatial distribution of replication centers throughout S phase in BrdUrd pulse labeled mouse fibroblasts.     

Automated detection of tunneling nanotubes in 3D images.

BACKGROUND: This paper presents an automated method for the identification of thin membrane tubes in 3D fluorescence images. These tubes, referred to as tunneling nanotubes (TNTs), are newly discovered intercellular structures that connect living cells through a membrane continuity. TNTs are 50-200 nm in diameter, crossing from one cell to another at their nearest distance. In microscopic images, they are seen as straight lines. It now emerges that the TNTs represent the underlying structure of a new type of cell-to-cell communication. METHODS: Our approach for the identification of TNTs is based on a combination of biological cell markers and known image processing techniques. Watershed segmentation and edge detectors are used to find cell borders, TNTs, and image artifacts. Mathematical morphology is employed at several stages of the processing chain. Two image channels are used for the calculations to improve classification of watershed regions into cells and background. One image channel displays cell borders and TNTs, the second is used for cell classification and displays the cytoplasmic compartments of the cells. The method for cell segmentation is 3D, and the TNT detection incorporates 3D information using various 2D projections. RESULTS: The TNT- and cell-detection were applied to numerous 3D stacks of images. A success rate of 67% was obtained compared with manual identification of the TNTs. The digitalized results were used to achieve statistical information of selected properties of TNTs. CONCLUSION: To further explore these structures, automated detection and quantification is desirable. Consequently, this automated recognition tool will be useful in biological studies on cell-to-cell communication where TNT quantification is essential.     

3-D Golgi and image analysis of the olfactory tubercle in schizophrenia

OBJECTIVE: To examine morphologic changes in the olfactory tubercle (OT) spiny neurons and astrocytes in schizophrenia (Sch) by means of quantitative 3-D Golgi and immunocytochemical studies. STUDY DESIGN: Free-floating vibrotome sections of postmortem brain tissue from 10 controls and 12 Sch cases were used for Golgi study and glial fibrillary acidic protein (GFAP) immunocytochemistry. A gray level image analysis was applied for quantitative estimation of GFAP-positive astrocytes on uniform, randomly sampled sections. This method is effective for low-contrast objects on an uneven background. Golgi-impregnated OT spiny neurons were analyzed both qualitatively and quantitatively in three dimensions with a semiautomated microscope-computer system. From digitized image of the neurons, various metric parameters were estimated to characterize the dendritic tree. RESULTS: In cases of Sch, degenerative changes in the dendrites of OT spiny neurons were revealed. A decrease in the maximal radius of the dendritic tree and total length of dendrites were accompanied by an increase in the length density of dendrites. Hypertrophy and a more-intensive GFAP reaction of astrocytes were found in OT of Sch. CONCLUSION: Based on these results, one can hypothesize that OT spiny neurons in Sch are involved in the process of dendritic reorganization, including degenerative changes in dendrites.     

A statistical approach to computer processing of cryo-electron microscope images: virion classification and 3-D reconstruction

The scattering density of the virus is represented as a truncated weighted sum of orthonormal basis functions in spherical coordinates, where the angular dependence of each basis function has icosahedral symmetry. A statistical model of the image formation process is proposed and the maximum likelihood estimation method computed by an expectation-maximization algorithm is used to estimate the weights in the sum and thereby compute a 3-D reconstruction of the virus particle. If multiple types of virus particle are represented in the boxed images then multiple 3-D reconstructions are computed simultaneously without first requiring that the type of particle shown in each boxed image be determined. Examples of the procedure are described for viruses with known structure: (1). 3-D reconstruction of Flockhouse Virus from experimental images, (2). 3-D reconstruction of the capsid of Nudaurelia Omega Capensis Virus from synthetic images, and (3). 3-D reconstruction of both the capsid and the procapsid of Nudaurelia Omega Capensis Virus from a mixture of unclassified synthetic images.     

An ab initio algorithm for low-resolution 3-D reconstructions from cryoelectron microscopy images

A statistical method for determining low-resolution 3-D reconstructions of virus particles from cryoelectron microscope images by an ab initio algorithm is described. The method begins with a novel linear reconstruction method that generates a spherically symmetric reconstruction, which is followed by a nonlinear reconstruction method implementing an expectation-maximization procedure using the spherically symmetric reconstruction as an initial condition and resulting in a reconstruction with icosahedral symmetry. An important characteristic of the complete method is that very little need be known about the particle before the reconstruction is computed, in particular, only the type of symmetry and inner and outer radii. The method is demonstrated on synthetic cowpea mosaic virus data, and its robustness to 5% errors in the contrast transfer function, 5% errors in the location of the center of the particles in the images, and 5% distortion in the 3-D structure from which the images are derived is demonstrated numerically.     

Automated selection of DAB-labeled tissue for immunohistochemical quantification.

The increased use of immunohistochemistry (IHC) in both clinical and basic research settings has led to the development of techniques for acquiring quantitative information from immunostains. Staining correlates with absolute protein levels and has been investigated as a clinical tool for patient diagnosis and prognosis. For these reasons, automated imaging methods have been developed in an attempt to standardize IHC analysis. We propose a novel imaging technique in which brightfield images of diaminobenzidene (DAB)-labeled antigens are converted to normalized blue images, allowing automated identification of positively stained tissue. A statistical analysis compared our method with seven previously published imaging techniques by measuring each one's agreement with manual analysis by two observers. Eighteen DAB-stained images showing a range of protein levels were used. Accuracy was assessed by calculating the percentage of pixels misclassified using each technique compared with a manual standard. Bland-Altman analysis was then used to show the extent to which misclassification affected staining quantification. Many of the techniques were inconsistent in classifying DAB staining due to background interference, but our method was statistically the most accurate and consistent across all staining levels.