Boaz Lerner

Toward a completely automatic neural-network-based human chromosome analysis

The application of neural networks (NNs) to automatic analysis of chromosome images is investigated in this paper. All aspects of the analysis, namely segmentation, feature description, selection and extraction, and classification, are studied. As part of the segmentation process, the separation of clusters of partially occluded chromosomes, which is the critical stage that state-of-the-art chromosome analyzers usually fail to accomplish, is performed. First, a moment representation of the image pixels is clustered to create a binary image without a need for threshold selection. Based on the binary image, lines connecting cut points imply possible separations. These hypotheses are verified by a multilayer perceptron (MLP) NN that classifies the two segments created by each separating line. Use of a classification-driven segmentation process gives very promising results without a need for shape modeling or an excessive use of heuristics. In addition, an NN implementation of Sammons mapping using principal component based initialization is applied to feature extraction, significantly reducing the dimensionality of the feature space and allowing high classification capability. Finally, by applying MLP based hierarchical classification strategies to a well-explored chromosome database, we achieve a classification performance of 83.6%. This is higher than ever published on this database and an improvement of more than 10% in the error rate. Therefore, basing a chromosome analysis on the NN-based techniques that are developed in this research leads toward a completely automatic human chromosome analysis.

Bayesian Network Structure Learning by Recursive Autonomy Identification

We propose the recursive autonomy identification (RAI) algorithm for constraint-based (CB) Bayesian network structure learning. The RAI algorithm learns the structure by sequential application of conditional independence (CI) tests, edge direction and structure decomposition into autonomous sub-structures. The sequence of operations is performed recursively for each autonomous sub-structure while simultaneously increasing the order of the CI test. While other CB algorithms d-separate structures and then direct the resulted undirected graph, the RAI algorithm combines the two processes from the outset and along the procedure. By this means and due to structure decomposition, learning a structure using RAI requires a smaller number of CI tests of high orders. This reduces the complexity and run-time of the algorithm and increases the accuracy by diminishing the curse-of-dimensionality. When the RAI algorithm learned structures from databases representing synthetic problems, known networks and natural problems, it demonstrated superiority with respect to computational complexity, run-time, structural correctness and classification accuracy over the PC, Three Phase Dependency Analysis, Optimal Reinsertion, greedy search, Greedy Equivalence Search, Sparse Candidate, and Max-Min Hill-Climbing algorithms.

On pattern classification with Sammon's nonlinear mapping an experimental study

Abstract Sammons mapping is conventionally used for exploratory data projection, and as such is usually inapplicable for classification. In this paper we apply a neural network (NN) implementation of Sammons mapping to classification by extracting an arbitrary number of projections. The projection map and classification accuracy of the mapping are compared with those of the auto-associative NN (AANN), multilayer perceptron (MLP) and principal component (PC) feature extractor for chromosome data. We demonstrate that chromosome classification based on Sammons (unsupervised) mapping is superior to the classification based on the AANN and PC feature extractor and highly comparable with that based on the (supervised) MLP. c 1998 Pattern Recognition Society.

Medial axis transform-based features and a neural network for human chromosome classification

Abstract Medial axis transform (MAT) based features and a multilayer perceptron (MLP) neural network (NN) were used for human chromosome classification. Two approaches to the MAT, one based on skeletonization and the other based on a piecewise linear (PWL) approximation, were examined. The former yielded a finer medial axis, as well as better chromosome classification performances. Geometrical along with intensity-based features were extracted and tested. The probability of correct training set classification of five chromosome types was 99.3–99.6%. The probability of correct test set classification was greater than 98% and greater than 97% using features extracted by the first and second approaches, respectively. It was found that only 5–10, out of all the considered features, were required to correctly classify the chromosomes with almost no performance degradation.

Support vector machine-based image classification for genetic syndrome diagnosis

We implement structural risk minimization and cross-validation in order to optimize kernel and parameters of a support vector machine (SVM) and multiclass SVM-based image classifiers, thereby enabling the diagnosis of genetic abnormalities. By thresholding the distance of patterns from the hypothesis separating the classes we reject a percentage of the miss-classified patterns reducing the expected risk. Accurate performance of the SVM in comparison to other state-of-the-art classifiers demonstrates the benefit of SVM-based genetic syndrome diagnosis.

Feature representation and signal classification in fluorescence in-situ hybridization image analysis

Fast and accurate analysis of fluorescence in-situ hybridization images for signal counting will depend mainly upon two components: a classifier to discriminate between artifacts and valid signals of several fluorophores (colors), and well discriminating features to represent the signals. Our previous work (2001) has focused on the first component. To investigate the second component, we evaluate candidate feature sets by illustrating the probability density functions and scatter plots for the features. The analysis provides first insight into dependencies between features, indicates the relative importance of members of a feature set, and helps in identifying sources of potential classification errors. Class separability yielded by different feature subsets is evaluated using the accuracy of several neural network-based classification strategies, some of them hierarchical, as well as using a feature selection technique making use of a scatter criterion. Although applied to cytogenetics, the paper presents a comprehensive, unifying methodology of qualitative and quantitative evaluation of pattern feature representation essential for accurate image classification. This methodology is applicable to many other real-world pattern recognition problems.

Accurate and Fast Off and Online Fuzzy ARTMAP-Based Image Classification With Application to Genetic Abnormality Diagnosis

We propose and investigate the fuzzy ARTMAP neural network in off and online classification of fluorescence in situ hybridization image signals enabling clinical diagnosis of numerical genetic abnormalities. We evaluate the classification task (detecting a several abnormalities separately or simultaneously), classifier paradigm (monolithic or hierarchical), ordering strategy for the training patterns (averaging or voting), training mode (for one epoch, with validation or until completion) and model sensitivity to parameters. We find the fuzzy ARTMAP accurate in accomplishing both tasks requiring only very few training epochs. Also, selecting a training ordering by voting is more precise than if averaging over orderings. If trained for only one epoch, the fuzzy ARTMAP provides fast, yet stable and accurate learning as well as insensitivity to model complexity. Early stop of training using a validation set reduces the fuzzy ARTMAP complexity as for other machine learning models but cannot improve accuracy beyond that achieved when training is completed. Compared to other machine learning models, the fuzzy ARTMAP does not loose but gain accuracy when overtrained, although increasing its number of categories. Learned incrementally, the fuzzy ARTMAP reaches its ultimate accuracy very fast obtaining most of its data representation capability and accuracy by using only a few examples. Finally, the fuzzy ARTMAP accuracy for this domain is comparable with those of the multilayer perceptron and support vector machine and superior to those of the naive Bayesian and linear classifiers

A classification-driven partially occluded object segmentation (CPOOS) method with application to chromosome analysis

Classification of segment images created by connecting points of high concavity along curvatures is used to resolve partial occlusion in images. Modeling of shape or curvature is not necessary nor is the traditional excessive use of heuristics. Applied to human cell images, 82.6% of the analyzed clusters of chromosomes are correctly separated, rising to 90.5% following rejection of 8.7% of the images.

Bayesian fluorescence in situ hybridisation signal classification

Previous research has indicated the significance of accurate classification of fluorescence in situ hybridisation (FISH) signals for the detection of genetic abnormalities. Based on well-discriminating features and a trainable neural network (NN) classifier, a previous system enabled highly-accurate classification of valid signals and artefacts of two fluorophores. However, since this system employed several features that are considered independent, the naive Bayesian classifier (NBC) is suggested here as an alternative to the NN. The NBC independence assumption permits the decomposition of the high-dimensional likelihood of the model for the data into a product of one-dimensional probability densities. The naive independence assumption together with the Bayesian methodology allow the NBC to predict a posteriori probabilities of class membership using estimated class-conditional densities in a close and simple form. Since the probability densities are the only parameters of the NBC, the misclassification rate of the model is determined exclusively by the quality of density estimation. Densities are evaluated by three methods: single Gaussian estimation (SGE; parametric method), Gaussian mixture model assuming spherical covariance matrices (GMM; semi-parametric method) and kernel density estimation (KDE; non-parametric method). For low-dimensional densities, the GMM generally outperforms the KDE that tends to overfit the training set at the cost of reduced generalisation capability. But, it is the GMM that loses some accuracy when modelling higher-dimensional densities due to the violation of the assumption of spherical covariance matrices when dependent features are added to the set. Compared with these two methods, the SGE and NN provide inferior and superior performance, respectively. However, the NBC avoids the intensive training and optimisation required for the NN, demanding extensive resources and experimentation. Therefore, when supporting these two classifiers, the system enables a trade-off between the NN performance and NBC simplicity of implementation.

Feature extraction by neural network nonlinear mapping for pattern classification

Feature extraction for exploratory data projection aims for data visualization by a projection of a high-dimensional space onto two or three-dimensional space, while feature extraction for classification generally requires more than two or three features. We study extraction of more than three features, using neural network (NN) implementation of Sammons mapping to be applied for classification. The experiments reveal that Sammons mapping, the multilayer perceptron (MLP) and the principal component analysis (PCA) based feature extractors yield similar classification performance. We investigate a random- and PCA-based initializations of Sammons mapping. When the PCA is applied to initialize Sammons projection, only one experiment is required and only a fraction of the training period is needed to achieve performance comparable with that of the random initialization. Furthermore, the PCA based initialization affords better human chromosome classification performance even when using a few eigenvectors.