Analysis of machine-selected cells with an image analysis system in normal and abnormal cervical specimens

An analysis has been performed of visual diagnostic criteria used in cervical cytology applied to machine selected cells in relation to automated classification based on variables, which can be recorded in an image system with automated cell search and segmentation, feature extraction and classification. A 98% accuracy could be obtained with the choice of the most ideal statistical methods for discrimination and the use of the most powerful variables recorded in the image system when compared with consensus of the visual diagnoses based on established cytological criteria for diagnosis of cancer and precancer of the cervix uteri. The most powerful discriminatory variables in the image system (of 17 recorded) for discrimination between normal and abnormal epithelial cells were, in addition to nuclear extinction, cytoplasmic extinction and cytoplasmic shape. It is concluded that the visual classification of cervical cells is highly accurate with experienced observers and that imaging microscopes can be trained to nearly equal this accuracy with appropriate statistical methods of discrimination. The problem of creating fully automated systems, however, also requires the inclusion of even more effective discriminatory variables and also the solution of such problems as automatic cell search, segmentation, artifact rejection, feature extraction, classification and electronic stability in order to become cost-effective.     

High-resolution and contextual analysis for the diagnosis of fine needle aspirates of breast.

A study was undertaken to confirm earlier work on a smaller number of patients that had suggested that medium-resolution contextual analysis complements high-resolution individual cell analysis for cytomorphometric classification of fine needle aspirate smears of breast. The objectives of this study were to improve and verify the method. Sixty-one biopsy-confirmed hematoxylin and eosin-stained aspirate smears of breast were restained using the Feulgen technique. Individual nuclei were digitized at a resolution of 0.25 micron. Features describing size, shape, density and texture were extracted from the images. Individual cell analysis correctly classified 84% of cases, contextual analysis correctly classified 70% of cases, and the combined use of both techniques resulted in 87% classification accuracy. However, if fibroadenoma cases are excluded, the combined correct classification rate is 93%. Geometric and densitometric features contributed most to correct classification in individual cell analysis, while the most important contextual feature was the number of clusters per scene. We conclude that the addition of quantitative measures of smear patterns, termed "contextual analysis," improves automated classification schemes.     

A Comparison of Supervised Machine Learning Algorithms and Feature Vectors for MS Lesion Segmentation Using Multimodal Structural MRI

Machine learning is a popular method for mining and analyzing large collections of medical data. We focus on a particular problem from medical research, supervised multiple sclerosis (MS) lesion segmentation in structural magnetic resonance imaging (MRI). We examine the extent to which the choice of machine learning or classification algorithm and feature extraction function impacts the performance of lesion segmentation methods. As quantitative measures derived from structural MRI are important clinical tools for research into the pathophysiology and natural history of MS, the development of automated lesion segmentation methods is an active research field. Yet, little is known about what drives performance of these methods. We evaluate the performance of automated MS lesion segmentation methods, which consist of a supervised classification algorithm composed with a feature extraction function. These feature extraction functions act on the observed T1-weighted (T1-w), T2-weighted (T2-w) and fluid-attenuated inversion recovery (FLAIR) MRI voxel intensities. Each MRI study has a manual lesion segmentation that we use to train and validate the supervised classification algorithms. Our main finding is that the differences in predictive performance are due more to differences in the feature vectors, rather than the machine learning or classification algorithms. Features that incorporate information from neighboring voxels in the brain were found to increase performance substantially. For lesion segmentation, we conclude that it is better to use simple, interpretable, and fast algorithms, such as logistic regression, linear discriminant analysis, and quadratic discriminant analysis, and to develop the features to improve performance.     

MBGD: microbial genome database for comparative analysis

MBGD is a workbench system for comparative analysis of completely sequenced microbial genomes. The central function of MBGD is to create an orthologous gene classification table using precomputed all-against-all similarity relationships among genes in multiple genomes. In MBGD, an automated classification algorithm has been implemented so that users can create their own classification table by specifying a set of organisms and parameters. This feature is especially useful when the user's interest is focused on some taxonomically related organisms. The created classification table is stored into the database and can be explored combining with the data of individual genomes as well as similarity relationships among genomes. Using these data, users can carry out comparative analyses from various points of view, such as phylogenetic pattern analysis, gene order comparison and detailed gene structure comparison. MBGD is accessible at http://mbgd.genome.ad.jp/.     

Automated comet assay analysis.

BACKGROUND: Recently the "comet assay" or "single-cell gel electrophoresis assay" has been established as a sensitive method for the detection of DNA damage and repair. Most of the software now available to quantify various parameters for DNA damage requires the interaction of a human observer. In this report, we describe an automated analysis system that is based on self-developed software and hardware and needs minimal human interaction. METHODS: The image analysis is divided into two parts: 1) automatic cell recognition and comet classification and 2) quantification of desired comet parameters. Image preprocessing, segmentation, and feature classification were developed with algorithms based on mathematical morphology. To enhance evaluation speed, we have introduced parallel processing of data under the Windows NT operating system (Microsoft Corporation, Redmond, WA). Use of an analogue real-time autofocus unit (B?cker et al.: Phys Med Biol 1997;42:1981-1992) allows for faster analysis. RESULTS: Our recognition software shows a sensitivity of 95.2% and a specificity of 92.7% when tested on test samples from routine work with DNA damage by low-dose radiation (0-2 Gy). The parallel hardware and software concept enables us to analyze 100 comets on one slide in less than 15 min. CONCLUSIONS: A comparison of measurements made on the same samples by manual and automated analysis systems revealed that there are no significant differences. The slope of the dose-response curves and the repair kinetics are very similar and demonstrate that automatic comet assay analysis is possible.     

Evaluation of contextual analysis for computer classification of cervical smears

A procedure for automated analysis of cervical smears has been implemented in an image cytometry system. Smears are described exclusively in terms of global and contextual information extracted by pattern-recognition algorithms and represented by a vector of proportions of cellular object types. Linear discriminant functions, based on a Fisher criterion, are derived to classify smears with a cross-section of diagnoses into two broad categories, normal and abnormal. Results obtained from 83 smears indicate 78% correct classification. In contrast to most automated systems, good classification results were obtained in normal smears with benign changes caused by inflammation and with postmenopausal atrophia and in abnormals with mild dysplasia. These findings suggest that contextual analysis may be sensitive to subtle changes in cellular morphology and to progressive patterns of dysplasia. When used with standard isolated cell analysis, contextual analysis may provide additional complementary information for automated cervical prescreening.     

Laboratory Test of an Automated Cell Analysis System For Cervical Screening

An automated cell analysis system (Autoplan-MIAC) for the early detection of precancerous lesions of the cervix was tested under semi-routine conditions in a clinical cytology laboratory. A set of 1500 specimens, highly enriched with abnormal cases, was analysed. Cervical scrapings were collected in suspension and processed by cytocentrifugation for microscopy. Two slides were prepared from each sample: one for staining according to Papanicolaou for the visual reference diagnosis and one for Feulgen staining for automated analysis. the specimens were evaluated in two ways: the first one, which is referred to as the automated machine classification system (AMC), classifies the specimens according to the number and ratio of selected objects (alarms) and is a fully automated system. the second system classifies the specimens after visual evaluation of the stored alarms as they are displayed on a TV monitor, and is designated the interactive machine classification system (IMC). the AMC results showed a false positive rate of 16.5% when the cut-off threshold was selected so that all 117 positively diagnosed specimens were classified ‘positive’ by the system. In that case 87.4% of the CINI and 96.9% of the CINII cases were AMC-positive. the IMC results showed a false positive rate of 2.5%, when 86.3% of the CIN I cases, 96.9% of the CIN II cases and all CIN III and invasive carcinoma cases were positively classified.     

Small bowel image classification using cross-co-occurrence matrices on wavelet domain

This paper presents a novel system to compute the automated classification of wireless capsule endoscope images. Classification is achieved by a classical statistical approach, but novel features are extracted from the wavelet domain and they contain both color and texture information. First, a shift-invariant discrete wavelet transform (SIDWT) is computed to ensure that the multiresolution feature extraction scheme is robust to shifts. The SIDWT expands the signal (in a shift-invariant way) over the basis functions which maximize information. Then cross-co-occurrence matrices of wavelet subbands are calculated and used to extract both texture and color information. Canonical discriminant analysis is utilized to reduce the feature space and then a simple 1D classifier with the leave one out method is used to automatically classify normal and abnormal small bowel images. A classification rate of 94.7% is achieved with a database of 75 images (41 normal and 34 abnormal cases). The high success rate could be attributed to the robust feature set which combines multiresolutional color and texture features, with shift, scale and semi-rotational invariance. This result is very promising and the method could be used in a computer-aided diagnosis system or a content-based image retrieval scheme.     

Progress report of the TUDAB project for automated cancer cell detection.

Two methods for high resolution cell image data acquisition are applied routinely. Cells are either scanned by a computer controlled fast scanning microscope photometer (SMP) or a TV-camera. The software system for digital image analysis was completely revised and implemented on the PR 330 minicomputer. The system contains codes for primary cell data acquisition, segmentation of cells, cell feature extraction and statistical cell analysis. With this system, SMP and TV scanned cell data bases of PAP stained cells in vaginal smears, grouped into several classes, have been built up. Each data base contains 34 primary features and 20 feature combinations for each cell. A linear discriminant analysis is applied routinely for cell classification. The present state of the system and its operation are described, cell features and classification results are shown, and future steps for a prescreening strategy are discussed.     

Adaptive frequency scaled wavelet packet decomposition for frog call classification

Environmental changes have put great pressure on biological systems leading to the rapid decline of biodiversity. To monitor this change and protect biodiversity, animal vocalizations have been widely explored by the aid of deploying acoustic sensors in the field. Consequently, large volumes of acoustic data are collected. However, traditional manual methods that require ecologists to physically visit sites to collect biodiversity data are both costly and time consuming. Therefore it is essential to develop new semi-automated and automated methods to identify species in automated audio recordings. In this study, a novel feature extraction method based on wavelet packet decomposition is proposed for frog call classification. After syllable segmentation, the advertisement call of each frog syllable is represented by a spectral peak track, from which track duration, dominant frequency and oscillation rate are calculated. Then, a k-means clustering algorithm is applied to the dominant frequency, and the centroids of clustering results are used to generate the frequency scale for wavelet packet decomposition (WPD). Next, a new feature set named adaptive frequency scaled wavelet packet decomposition sub-band cepstral coefficients is extracted by performing WPD on the windowed frog calls. Furthermore, the statistics of all feature vectors over each windowed signal are calculated for producing the final feature set. Finally, two well-known classifiers, a k-nearest neighbour classifier and a support vector machine classifier, are used for classification. In our experiments, we use two different datasets from Queensland, Australia (18 frog species from commercial recordings and field recordings of 8 frog species from James Cook University recordings). The weighted classification accuracy with our proposed method is 99.5% and 97.4% for 18 frog species and 8 frog species respectively, which outperforms all other comparable methods.     

Tissue counter analysis of histologic sections of melanoma: influence of mask size and shape, feature selection, statistical methods and tissue preparation.

BACKGROUND: Tissue counter analysis is an image analysis tool designed for the detection of structures in complex images at the macroscopic or microscopic scale. As a basic principle, small square or circular measuring masks are randomly placed across the image and image analysis parameters are obtained for each mask. Based on learning sets, statistical classification procedures are generated which facilitate an automated classification of new data sets. OBJECTIVE: To evaluate the influence of the size and shape of the measuring masks as well as the importance of feature selection, statistical procedures and technical preparation of slides on the performance of tissue counter analysis in microscopic images. As main quality measure of the final classification procedure, the percentage of elements that were correctly classified was used. STUDY DESIGN: HE-stained slides of 25 primary cutaneous melanomas were evaluated by tissue counter analysis for the recognition of melanoma elements (section area occupied by tumour cells) in contrast to other tissue elements and background elements. Circular and square measuring masks, various subsets of image analysis features and classification and regression trees compared with linear discriminant analysis as statistical alternatives were used. The percentage of elements that were correctly classified by the various classification procedures was assessed. In order to evaluate the applicability to slides obtained from different laboratories, the best procedure was automatically applied in a test set of another 50 cases of primary melanoma derived from the same laboratory as the learning set and two test sets of 20 cases each derived from two different laboratories, and the measurements of melanoma area in these cases were compared with conventional assessment of vertical tumour thickness. RESULTS: Square measuring masks were slightly superior to circular masks, and larger masks (64 or 128 pixels in diameter) were superior to smaller masks (8 to 32 pixels in diameter). As far as the subsets of image analysis features were concerned, colour features were superior to densitometric and Haralick texture features. Statistical moments of the grey level distribution were of least significance. CART (classification and regression tree) analysis turned out to be superior to linear discriminant analysis. In the best setting, 95% of melanoma tissue elements were correctly recognized. Automated measurement of melanoma area in the independent test sets yielded a correlation of r=0.846 with vertical tumour thickness (p<0.001), similar to the relationship reported for manual measurements. The test sets obtained from different laboratories yielded comparable results. CONCLUSIONS: Large, square measuring masks, colour features and CART analysis provide a useful setting for the automated measurement of melanoma tissue in tissue counter analysis, which can also be used for slides derived from different laboratories.     

Comparison of Papanicolaou staining and Feulgen staining for automated prescreening

Feulgen staining is considered to be a quantitative DNA-specific cytochemical procedure. The applicability of this staining in high-resolution cytometry was tested in comparison with a regressive Papanicolaou staining. Papanicolaou-stained or Feulgen-stained intermediate and carcinoma cells selected by a cytologist were examined with a Zeiss scanning microscope photometer at 546 and 560 nm, respectively. After cell image segmentation and feature extraction, a statistical data evaluation was carried out by computer. Cell distributions with respect to four selected nuclear features demonstrated the influence of the staining procedure on cell feature measurements. The discriminatory power of the classification system as related to both staining procedures was studied using discriminant analysis. Using only nuclear features, a 7.3% improvement of the overall correct classification rate (from 85.0% to 92.3%) was achieved using Feulgen staining. The misclassification rate was simultaneously reduced by 50%. Using cytoplasmic as well as nuclear features, a 98% rate of correct classification was achieved.     

Automated Analysis of a Diverse Synapse Population

Synapses of the mammalian central nervous system are highly diverse in function and molecular composition. Synapse diversity per se may be critical to brain function, since memory and homeostatic mechanisms are thought to be rooted primarily in activity-dependent plastic changes in specific subsets of individual synapses. Unfortunately, the measurement of synapse diversity has been restricted by the limitations of methods capable of measuring synapse properties at the level of individual synapses. Array tomography is a new high-resolution, high-throughput proteomic imaging method that has the potential to advance the measurement of unit-level synapse diversity across large and diverse synapse populations. Here we present an automated feature extraction and classification algorithm designed to quantify synapses from high-dimensional array tomographic data too voluminous for manual analysis. We demonstrate the use of this method to quantify laminar distributions of synapses in mouse somatosensory cortex and validate the classification process by detecting the presence of known but uncommon proteomic profiles. Such classification and quantification will be highly useful in identifying specific subpopulations of synapses exhibiting plasticity in response to perturbations from the environment or the sensory periphery.     

Rapid 3D phenotypic analysis of neurons and organoids using data-driven cell segmentation-free machine learning

Phenotypic profiling of large three-dimensional microscopy data sets has not been widely adopted due to the challenges posed by cell segmentation and feature selection. The computational demands of automated processing further limit analysis of hard-to-segment images such as of neurons and organoids. Here we describe a comprehensive shallow-learning framework for automated quantitative phenotyping of three-dimensional (3D) image data using unsupervised data-driven voxel-based feature learning, which enables computationally facile classification, clustering and advanced data visualization. We demonstrate the analysis potential on complex 3D images by investigating the phenotypic alterations of: neurons in response to apoptosis-inducing treatments and morphogenesis for oncogene-expressing human mammary gland acinar organoids. Our novel implementation of image analysis algorithms called Phindr3D allowed rapid implementation of data-driven voxel-based feature learning into 3D high content analysis (HCA) operations and constitutes a major practical advance as the computed assignments represent the biology while preserving the heterogeneity of the underlying data. Phindr3D is provided as Matlab code and as a stand-alone program (https://github.com/DWALab/Phindr3D).     

Artificial intelligence in automated classification of rat vaginal smear cells.

Microscopic examination of vaginal smears has been used routinely to determine the stage of the estrous cycle of female rats in reproductive research. The stage of the estrous cycle is based on relative counts of nucleated epithelial cells, cornified epithelial cells and leukocytes. The purpose of this project was to explore automation of vaginal smear analysis using image processing and artificial intelligence techniques. A fully connected back-propagation neural network was used to locate all potential objects in a digitized scene. A unique algorithm was then employed to center a subsequent sampling box to collect pixel intensity values from the red and green components of each image. A final neural network was used in the classification of cell type. Neural networks were used because of their ability to generalize among input patterns and to tolerate extraneous noise due to variations in staining artifacts and aberrant illumination of the microscope field. This preliminary cell diagnosing system not only provides the basis for the fully automated system but also provides a method by which many other cytologic image processing problems can be automated.     

Influence of staining on fast automated cell segmentation, feature extraction and cell image analysis

FAZYTAN, a system for fast automated cell segmentation, cell image analysis and extraction of nuclear features, was used to analyze cervical cell images variously stained by the conventional Papanicolaou stain, the new Papanicolaou stain and hematoxylin and thionin only; the last two dyes are used as the nuclear stains in the two versions of the Papanicolaou stain. Other dyes were also tried in cell classification experiments. All cell images in the variously stained samples could be described by the same nuclear features as had been adapted for the discrimination of conventional-Papanicolaou-stained cells. Variances were lower for thionin-stained cells as compared with hematoxylin-stained cells. By application of spectrophotometry, it was confirmed that the spectra of the cytoplasmic counterstains are superimposed on those of the nuclear stains. It appears that a variety of dyes are suitable as cytologic stains for cell classification by the FAZYTAN system, provided that they achieve sufficiently strong nuclear-cytoplasmic contrast by precisely delineating the chromatin texture.