DNA Copy Number Selection Using Robust Structured Sparsity-Inducing Norms.

Array comparative genomic hybridization (aCGH) is a newly introduced method for the detection of copy number abnormalities associated with human diseases with special focus on cancer. Specific patterns in DNA copy number variations (CNVs) can be associated with certain disease types and can facilitate prognosis and progress monitoring of the disease. Machine learning techniques have been used to model the problem of tissue typing as a classification problem. Feature selection is an important part of the classification process, because many biological features are not related to the diseases and confuse the classification tasks. Multiple feature selection methods have been proposed in the different domains where classification has been applied. In this work, we will present a new feature selection method based on structured sparsity-inducing norms to identify the informative aCGH biomarkers which can help us classify different disease subtypes. To validate the performance of the proposed method, we experimentally compare it with existing feature selection methods on four publicly available aCGH data sets. In all empirical results, the proposed sparse learning based feature selection method consistently outperforms other related approaches. More important, we carefully investigate the aCGH biomarkers selected by our method, and the biological evidences in literature strongly support our results.

A web ontology for brain trauma patient computer-assisted rehabilitation.

In this paper we describe CABROnto, which is a web ontology for the semantic representation of the computer assisted brain trauma rehabilitation. This is a novel and emerging domain, since it employs the use of robotic devices, adaptation software and machine learning to facilitate interactive and adaptive rehabilitation care. We used Protégé 4.2 and Protégé-Owl schema editor. The primary goal of this ontology is to enable the reuse of the domain knowledge. CABROnto has nine main classes, more than 50 subclasses, existential and cardinality restrictions. The ontology can be found online at Bioportal.

Heterogeneous data fusion for brain tumor classification.

Current research in biomedical informatics involves analysis of multiple heterogeneous data sets. This includes patient demographics, clinical and pathology data, treatment history, patient outcomes as well as gene expression, DNA sequences and other information sources such as gene ontology. Analysis of these data sets could lead to better disease diagnosis, prognosis, treatment and drug discovery. In this report, we present a novel machine learning framework for brain tumor classification based on heterogeneous data fusion of metabolic and molecular datasets, including state-of-the-art high-resolution magic angle spinning (HRMAS) proton (1H) magnetic resonance spectroscopy and gene transcriptome profiling, obtained from intact brain tumor biopsies. Our experimental results show that our novel framework outperforms any analysis using individual dataset.

A novel surface registration algorithm with biomedical modeling applications.

In this paper, we propose a novel surface matching algorithm for arbitrarily shaped but simply connected 3-D objects. The spherical harmonic (SPHARM) method is used to describe these 3-D objects, and a novel surface registration approach is presented. The proposed technique is applied to various applications of medical image analysis. The results are compared with those using the traditional method, in which the first-order ellipsoid is used for establishing surface correspondence and aligning objects. In these applications, our surface alignment method is demonstrated to be more accurate and flexible than the traditional approach. This is due in large part to the fact that a new surface parameterization is generated by a shortcut that employs a useful rotational property of spherical harmonic basis functions for a fast implementation. In order to achieve a suitable computational speed for practical applications, we propose a fast alignment algorithm that improves computational complexity of the new surface registration method from O(n3) to O(n2).

Cardiac motion analysis to improve pacing site selection in CRT.

RATIONALE AND OBJECTIVES: The aim of the study is to build cardiac wall motion models to characterize mechanical dyssynchrony and predict pacing sites for the left ventricle of the heart in cardiac resynchronization therapy (CRT). MATERIALS AND METHODS: Cardiac magnetic resonance imaging data from 20 patients are used, in which half have heart failure problems. We propose two spatio-temporal ventricular motion models to analyze the mechanical dyssynchrony of heart: radial motion series and wall motion series (a time series of radial length or wall thickness change). The hierarchical agglomerative clustering technique is applied to the motion series to find candidate pacing sites. All experiments are performed separately on each ventricular motion model to facilitate performance comparison among models. RESULTS: The experimental results demonstrate that the proposed methods perform as well as we expect. Our techniques not only effectively generate the candidate pacing sites list that can help guide CRT, but also derive clustering results that can distinguish the heart conditions between patients and normals perfectly to help medical diagnosis and prognosis. After comparing the results between two different ventricular motion models, the wall motion series model shows a better performance. CONCLUSION: In a traditional CRT device deployment, pacing sites are selected without efficient prediction, which runs the risk of suboptimal benefits. Our techniques can extract useful wall motion information from ventricular mechanical dyssynchrony and identify the candidate pacing sites with maximum contraction delay to assist pacemaker implantation in CRT.

Overview: Computational analysis and decision support systems in oncology.

Computational analysis tools and decision support systems have increased their penetration in the support of clinical processes and management of medical data and knowledge. Applications range from adjunct tools for diagnosis and disease investigation to the treatment and monitoring of therapeutic procedures. As all medical fields, the field of oncology is affected. This special issue includes studies presenting research and applications of computational intelligence in oncology, covering four main areas: i) decision support systems (DSS) and artificial intelligence (AI) applications in oncology; ii) design and assessment of classification tools in oncology; iii) intelligent accessing, retrieving, and storing of medical images; and iv) intelligent telemedicine and telehealth applications in oncology.

Tumor classification based on DNA copy number aberrations determined using SNP arrays.

High-density single nucleotide polymorphism (SNP) array is a recently introduced technology that genotypes more than 10,000 human SNPs on a single array. It has been shown that SNP arrays can be used to determine not only SNP genotype calls, but also DNA copy number (DCN) aberrations, which are common in solid tumors. In the past, effective cancer classification has been demonstrated using microarray gene expression data, or DCN data derived from comparative genomic hybridization (CGH) arrays. However, the feasibility of cancer classification based on DCN aberrations determined by SNP arrays has not been previously investigated. In this study, we address this issue by applying state-of-the-art classification algorithms and feature selection algorithms to the DCN aberration data derived from a public SNP array dataset. Performance was measured via leave-one-out cross-validation (LOOCV) classification accuracy. Experimental results showed that the maximum accuracy was 73.33%, which is comparable to the maximum accuracy of 76.5% based on CGH-derived DCN data reported previously in the literature. These results suggest that DCN aberration data derived from SNP arrays is useful for etiology-based tumor classification.

Spatio-temporal modeling of lung images for cancer detection.

Perfusion magnetic resonance imaging (pMRI) is an important tool in assessing tumor angiogenesis for the early detection of lung cancer. This study presents a novel integrated framework for spatio-temporal modeling of pulmonary nodules in pMRI image sequences. After localizing a nodule region in each image, we perform segmentation in the region to extract the nodule boundary, then use thin-plate spline interpolation for nodule registration along the temporal dimension. The resulting spatio-temporal model can lead to many types of nodule characterization, e.g. a time-intensity profile of a nodule region, and be used to capture important angiogenic patterns in the lung that can distinguish between cancer and benign nodules and assist in early detection.

Modeling Time-Intensity Profiles for Pulmonary Nodules in MR Images.

Perfusion magnetic resonance imaging (pMRI) is an important tool to assess tumor angiogenesis for the early detection of lung cancer. This paper presents a novel integrated framework for spatio-temporal modeling of pulmonary nodules in pMRI image sequences. After localizing a nodule region in each image, we perform segmentation in the region to extract nodule boundary, and then use thin-plate spline interpolation for nodule registration along the temporal dimension. The resulting spatio-temporal model can lead to many types of nodule characterization. Time intensity profiles of nodules region capture important angiogenic patterns in the lung that can distinguish between cancer and benign nodules and help early detection.

Surface alignment of 3D spherical harmonic models: application to cardiac MRI analysis.

The spherical harmonic (SPHARM) description is a powerful surface modeling technique that can model arbitrarily shaped but simply connected 3D objects and has been used in many applications in medical imaging. Previous SPHARM techniques use the first order ellipsoid for establishing surface correspondence and aligning objects. However, this first order information may not be sufficient in many cases; a more general method for establishing surface correspondence would be to minimize the mean squared distance between two corresponding surfaces. In this paper, a new surface matching algorithm is proposed for 3D SPHARM models to achieve this goal. This algorithm employs a useful rotational property of spherical harmonic basis functions for a fast implementation. Applications of medical image analysis (e.g., spatio-temporal modeling of heart shape changes) are used to demonstrate this approach. Theoretical proofs and experimental results show that our approach is an accurate and flexible surface correspondence alignment method.

A prediction framework for cardiac resynchronization therapy via 4D cardiac motion analysis.

We propose a novel framework to predict pacing sites in the left ventricle (LV) of a heart and its result can be used to assist pacemaker implantation and programming in cardiac resynchronization therapy (CRT), a widely adopted therapy for heart failure patients. In a traditional CRT device deployment, pacing sites are selected without quantitative prediction. That runs the risk of suboptimal benefits. In this work, the spherical harmonic (SPHARM) description is employed to model the ventricular surfaces and a novel SPHARM-based surface correspondence approach is proposed to capture the ventricular wall motion. A hierarchical agglomerative clustering technique is applied to the time series of regional wall thickness to identify candidate pacing sites. Using clinical MRI data in our experiments, we demonstrate that the proposed framework can not only effectively identify suitable pacing sites, but also distinguish patients from normal subjects perfectly to help medical diagnosis and prognosis.

HykGene: a hybrid approach for selecting marker genes for phenotype classification using microarray gene expression data.

MOTIVATION: Recent studies have shown that microarray gene expression data are useful for phenotype classification of many diseases. A major problem in this classification is that the number of features (genes) greatly exceeds the number of instances (tissue samples). It has been shown that selecting a small set of informative genes can lead to improved classification accuracy. Many approaches have been proposed for this gene selection problem. Most of the previous gene ranking methods typically select 50-200 top-ranked genes and these genes are often highly correlated. Our goal is to select a small set of non-redundant marker genes that are most relevant for the classification task. RESULTS: To achieve this goal, we developed a novel hybrid approach that combines gene ranking and clustering analysis. In this approach, we first applied feature filtering algorithms to select a set of top-ranked genes, and then applied hierarchical clustering on these genes to generate a dendrogram. Finally, the dendrogram was analyzed by a sweep-line algorithm and marker genes are selected by collapsing dense clusters. Empirical study using three public datasets shows that our approach is capable of selecting relatively few marker genes while offering the same or better leave-one-out cross-validation accuracy compared with approaches that use top-ranked genes directly for classification. AVAILABILITY: The HykGene software is freely available at http://www.cs.dartmouth.edu/~wyh/software.htm CONTACT: wyh@cs.dartmouth.edu SUPPLEMENTARY INFORMATION: Supplementary material is available from http://www.cs.dartmouth.edu/~wyh/hykgene/supplement/index.htm.

A bipartite graph matching framework for finding correspondences between structural elements in two proteins.

A protein molecule consists one or more chains of amino acid sequences that fold into a complex three-dimensional structure. A protein's functions are often determined by its 3D structure, and so comparing the similarity of 3D structures between proteins is an important problem. To accomplish such comparison, one must align two proteins properly with rotation and translation in 3D space. Finding the correspondences between structural elements in the two proteins is the key step in many protein structure alignment algorithms. We introduce a new graph theoretic framework based on bipartite graph matching for finding sufficiently good correspondences. It is capable of providing both sequence-dependent and sequence-independent correspondences. It is a general framework for pair-wise matching of atoms, amino acids residues or secondary structure elements.

An extensible framework for sharing clinical guidelines and services.

Accurate and descriptive information from clinical studies guides improvements in health care. Clinical guidelines established by authoritative medical organizations provide such information in a standard form for medical professionals' reference. Previous work on electronically sharing clinical guidelines focuses on the idea of building unified clinical terminologies and sharing resources through centralized data repositories. In this paper we propose a novel five-layer framework called the Extensible Clinical Guidelines and Services Sharing Architecture (ECGSSA). This framework provides for clinical guideline sharing among autonomous service providers over a distributed architecture. Requests for exchange of guidelines are disseminated through Web Services through a registry mechanism. Currently we have adopted the Guideline Interchange Format (GLIF) from InterMed as the representation format and use the Open Grid Services Architecture (OGSA) to attain virtual organization of shared guideline and service resources. This approach will allow more flexibility for medical professionals to exchange their practice guidelines in an effort to improve quality of health care. Also, it extends the possibility of solving clinic-related computational problems through collaborative sharing of analytical services. A sample scenario is presented to explain the application of ECGSSA in distributed task assignment and service matching in medical image processing.

Efficient similarity retrieval for temporal shape sequences: a case study using cardiac MR images.

The spherical harmonics (SPHARM) approach has been used for the representation of shapes in many types of biomedical image data. We propose a SPHARM-based similarity comparison for shape sequences that allows fast similarity searches for dynamic objects and demonstrate it using 3D images of a beating heart. By using spherical harmonics to extract a small number of features that represent cardiac shape in each sequential state, we enable indexing and pruning of database entries with a multidimensional index tree (e.g. R*-tree) for fast retrieval. Our approach relies on obtaining selected landmarks to allow normalization within and between sequences. This framework is extensible to other application domains.

BP-neural network based-characterization of electrographic magnetohydrodynamic signals in MR.

Electrocardiographs (ECG) signal collected during magnetic resonance (MR) imaging is affected by signal artifact because magnetic fields produce competing signals, from moving conductors in the large vessels. That is called the magnetohydrodynamic effect, which makes it difficult to recognize ST-T changes from ECG signal collected in a magnetic field (MRI). Resolving that problem is important both for accurate triggering (elimination of false triggers from tall peaked T waves) and for monitoring (identifying if or when patient develops ischemia or myocardial injury). This work describes an algorithm based on neural network that is designed to cancel this artifact for ECG signal acquired during MR imaging.

An adaptive approach for image subtraction.

Image subtraction is widely used in angiography as a means of highlighting differences induced by contrast agents. New knowledge of previously unsuspected causes of disease, in particular, secondhand smoke exposure, spurs interest in pushing the limits of early accurate diagnosis. Simple image subtraction induces artifacts causing problems for ensuing measurements and 3D reconstruction. Image registration techniques have been used to partially solve this problem. However, a complete registration is slow, and misregistration often occurs in images where bones are surrounded by vessels with similar image characteristics. We propose an approach based on the idea of global match followed by local refinements. In the global match, an image pair is aligned using a similarity measure so as to reduce overall difference. In the local refinements, localized displacements and deformations of tissue are handled by a combination of techniques: image registration, region growing, erosion, and dilation. This approach is fast compared to registration based image subtraction and it can find vessels abutting a bone. It is designed to be especially suitable for large cross-section image stacks. With additional vessel connectivity analysis between adjacent slices, the algorithm provides a good foundation for 3D vessel reconstruction.