Genome-wide variant-based study of genetic effects with the largest neuroanatomic coverage.

BACKGROUND: Brain image genetics provides enormous opportunities for examining the effects of genetic variations on the brain. Many studies have shown that the structure, function, and abnormality (e.g., those related to Alzheimer's disease) of the brain are heritable. However, which genetic variations contribute to these phenotypic changes is not completely clear. Advances in neuroimaging and genetics have led us to obtain detailed brain anatomy and genome-wide information. These data offer us new opportunities to identify genetic variations such as single nucleotide polymorphisms (SNPs) that affect brain structure. In this paper, we perform a genome-wide variant-based study, and aim to identify top SNPs or SNP sets which have genetic effects with the largest neuroanotomic coverage at both voxel and region-of-interest (ROI) levels. Based on the voxelwise genome-wide association study (GWAS) results, we used the exhaustive search to find the top SNPs or SNP sets that have the largest voxel-based or ROI-based neuroanatomic coverage. For SNP sets with >2 SNPs, we proposed an efficient genetic algorithm to identify top SNP sets that can cover all ROIs or a specific ROI. RESULTS: We identified an ensemble of top SNPs, SNP-pairs and SNP-sets, whose effects have the largest neuroanatomic coverage. Experimental results on real imaging genetics data show that the proposed genetic algorithm is superior to the exhaustive search in terms of computational time for identifying top SNP-sets. CONCLUSIONS: We proposed and applied an informatics strategy to identify top SNPs, SNP-pairs and SNP-sets that have genetic effects with the largest neuroanatomic coverage. The proposed genetic algorithm offers an efficient solution to accomplish the task, especially for identifying top SNP-sets.

Decoding the Role of Secondary Motor Cortex Neuronal Ensembles during Cocaine Self-Administration: Insights from Longitudinal in vivo Calcium Imaging via Miniscopes.

Recent findings in our lab demonstrated that the risk of cocaine relapse is closely linked to the hyperexcitability of cortical pyramidal neurons in the secondary motor cortex (M2), noticeable 45 days after cocaine intravenous self-administration (IVSA). The present study was designed to explore the underlying mechanisms of neuronal alterations in M2. Our hypothesis was that M2 neurons were affected directly by cocaine taking behaviors. This hypothesis was tested by monitoring individual neuronal activity in M2 using MiniScopes for in vivo Ca 2+ imaging in C57BL/6J mice when they had access to cocaine IVSA as a reinforcement (RNF) contingent to active lever press (ALP) but not to inactive lever press (ILP). With support of our established pipeline to processing Ca 2+ imaging data, the current study was designed to monitor M2 neuronal ensembles at the single-neuron level in real time with high temporal resolution and high throughput in each IVSA session and longitudinally among multiple IVSA sessions. Specifically, five consecutive 1-hr daily IVSA sessions were used to model the initial cocaine taking behaviors. Besides detailed analyses of IVSA events (ALP, ILP, and RNF), the data from Ca 2+ imaging recordings in M2 were analyzed by (1) comparing neuronal activation within a daily IVSA session (i.e., the first vs. the last 15 min) and between different daily sessions (i.e., the first vs. the last IVSA day), (2) associating Ca 2+ transients with individual IVSA events, and (3) correlating Ca 2+ transients with the cumulative effects of IVSA events. Our data demonstrated that M2 neurons are exquisitely sensitive to and significantly affected by concurrent operant behaviors and the history of drug exposure, which in turn sculpt the upcoming operant behaviors and the response to drugs. As critical nodes of the reward loop, M2 neurons appear to be the governing center orchestrating the establishment of addiction-like behaviors.

Hippocampal Subregion and Gene Detection in Alzheimer's Disease Based on Genetic Clustering Random Forest.

The distinguishable subregions that compose the hippocampus are differently involved in functions associated with Alzheimer's disease (AD). Thus, the identification of hippocampal subregions and genes that classify AD and healthy control (HC) groups with high accuracy is meaningful. In this study, by jointly analyzing the multimodal data, we propose a novel method to construct fusion features and a classification method based on the random forest for identifying the important features. Specifically, we construct the fusion features using the gene sequence and subregions correlation to reduce the diversity in same group. Moreover, samples and features are selected randomly to construct a random forest, and genetic algorithm and clustering evolutionary are used to amplify the difference in initial decision trees and evolve the trees. The features in resulting decision trees that reach the peak classification are the important "subregion gene pairs". The findings verify that our method outperforms well in classification performance and generalization. Particularly, we identified some significant subregions and genes, such as hippocampus amygdala transition area (HATA), fimbria, parasubiculum and genes included RYR3 and PRKCE. These discoveries provide some new candidate genes for AD and demonstrate the contribution of hippocampal subregions and genes to AD.

Quantification of tooth displacement from cone-beam computed tomography images.

INTRODUCTION: The objectives of this study were to demonstrate a method that could be used to quantify three-dimensional (3D) tooth displacement from cone-beam computed tomography (CBCT) images and to assess its accuracy. METHODS: Images of the same mandible taken 2 weeks apart with no treatment were used. Four mandibular teeth-left lateral incisor, left canine, left first premolar, and left first molar-either remained unmoved or were artificially displaced with known values on 1 image to simulate after-treatment conditions. The iterative closest point method was used to superimpose the unchanged bony part of the mandible and to find the transformation matrix between a tooth's 2 positions, before and after displacement. Tooth displacement was calculated from the transformation matrix. RESULTS: All 6 displacement components in terms of translations along and rotations about the 3 axes on the tooth were obtained. The results showed that the errors could be managed: they were less than 5% in translation and 10% in rotation. CONCLUSIONS: The 3D tooth displacement can be obtained from CBCT images, and the accuracy is acceptable for clinical use and can be improved when the quality of the images improves.

Interactive Machine Learning by Visualization: A Small Data Solution.

Machine learning algorithms and traditional data mining process usually require a large volume of data to train the algorithm-specific models, with little or no user feedback during the model building process. Such a "big data" based automatic learning strategy is sometimes unrealistic for applications where data collection or processing is very expensive or difficult, such as in clinical trials. Furthermore, expert knowledge can be very valuable in the model building process in some fields such as biomedical sciences. In this paper, we propose a new visual analytics approach to interactive machine learning and visual data mining. In this approach, multi-dimensional data visualization techniques are employed to facilitate user interactions with the machine learning and mining process. This allows dynamic user feedback in different forms, such as data selection, data labeling, and data correction, to enhance the efficiency of model building. In particular, this approach can significantly reduce the amount of data required for training an accurate model, and therefore can be highly impactful for applications where large amount of data is hard to obtain. The proposed approach is tested on two application problems: the handwriting recognition (classification) problem and the human cognitive score prediction (regression) problem. Both experiments show that visualization supported interactive machine learning and data mining can achieve the same accuracy as an automatic process can with much smaller training data sets.

Heritability Estimation of Reliable Connectomic Features.

Brain imaging genetics is an emerging research field to explore the underlying genetic architecture of brain structure and function measured by different imaging modalities. However, not all the changes in the brain are a consequential result of genetic effect and it is usually unknown which imaging phenotypes are promising for genetic analyses. In this paper, we focus on identifying highly heritable measures of structural brain networks derived from diffusion weighted imaging data. Using the twin data from the Human Connectome Project (HCP), we evaluated the reliability of fractional anisotropy measure, fiber length and fiber number of each edge in the structural connectome and seven network level measures using intraclass correlation coefficients. We then estimated the heritability of those reliable network measures using SOLAR-Eclipse software. Across all 64,620 network edges between 360 brain regions in the Glasser parcellation, we observed ~5% of them with significantly high heritability in fractional anisotropy, fiber length or fiber number. All the tested network level measures, capturing the network integrality, segregation or resilience, are highly heritable, with variance explained by the additive genetic effect ranging from 59% to 77%.

JOINT EXPLORATION AND MINING OF MEMORY-RELEVANT BRAIN ANATOMIC AND CONNECTOMIC PATTERNS VIA A THREE-WAY ASSOCIATION MODEL.

Early change in memory performance is a key symptom of many brain diseases, but its underlying mechanism remains largely unknown. While structural MRI has been playing an essential role in revealing potentially relevant brain regions, increasing availability of diffusion MRI data (e.g., Human Connectome Project (HCP)) provides excellent opportunities for exploration of their complex coordination. Given the complementary information held in these two imaging modalities, we hypothesize that studying them as a whole, rather than individually, and exploring their association will provide us valuable insights of the memory mechanism. However, many existing association methods, such as sparse canonical correlation analysis (SCCA), only manage to handle two-way association and thus cannot guarantee the selection of biomarkers and associations to be memory relevant. To overcome this limitation, we propose a new outcome-relevant SCCA model (OSCCA) together with a new algorithm to enable the three-way associations among brain connectivity, anatomic structure and episodic memory performance. In comparison with traditional SCCA, we demonstrate the effectiveness of our model with both synthetic and real data from the HCP cohort.

GPU Accelerated Browser for Neuroimaging Genomics.

Neuroimaging genomics is an emerging field that provides exciting opportunities to understand the genetic basis of brain structure and function. The unprecedented scale and complexity of the imaging and genomics data, however, have presented critical computational bottlenecks. In this work we present our initial efforts towards building an interactive visual exploratory system for mining big data in neuroimaging genomics. A GPU accelerated browsing tool for neuroimaging genomics is created that implements the ANOVA algorithm for single nucleotide polymorphism (SNP) based analysis and the VEGAS algorithm for gene-based analysis, and executes them at interactive rates. The ANOVA algorithm is 110 times faster than the 4-core OpenMP version, while the VEGAS algorithm is 375 times faster than its 4-core OpenMP counter part. This approach lays a solid foundation for researchers to address the challenges of mining large-scale imaging genomics datasets via interactive visual exploration.

Iterative volume morphing and learning for mobile tumor based on 4DCT.

During image-guided cancer radiation treatment, three-dimensional (3D) tumor volumetric information is important for treatment success. However, it is typically not feasible to image a patient's 3D tumor continuously in real time during treatment due to concern over excessive patient radiation dose. We present a new iterative morphing algorithm to predict the real-time 3D tumor volume based on time-resolved computed tomography (4DCT) acquired before treatment. An offline iterative learning process has been designed to derive a target volumetric deformation function from one breathing phase to another. Real-time volumetric prediction is performed to derive the target 3D volume during treatment delivery. The proposed iterative deformable approach for tumor volume morphing and prediction based on 4DCT is innovative because it makes three major contributions: (1) a novel approach to landmark selection on 3D tumor surfaces using a minimum bounding box; (2) an iterative morphing algorithm to generate the 3D tumor volume using mapped landmarks; and (3) an online tumor volume prediction strategy based on previously trained deformation functions utilizing 4DCT. The experimental performance showed that the maximum morphing deviations are 0.27% and 1.25% for original patient data and artificially generated data, which is promising. This newly developed algorithm and implementation will have important applications for treatment planning, dose calculation and treatment validation in cancer radiation treatment.

Brain explorer for connectomic analysis.

Visualization plays a vital role in the analysis of multimodal neuroimaging data. A major challenge in neuroimaging visualization is how to integrate structural, functional, and connectivity data to form a comprehensive visual context for data exploration, quality control, and hypothesis discovery. We develop a new integrated visualization solution for brain imaging data by combining scientific and information visualization techniques within the context of the same anatomical structure. In this paper, new surface texture techniques are developed to map non-spatial attributes onto both 3D brain surfaces and a planar volume map which is generated by the proposed volume rendering technique, spherical volume rendering. Two types of non-spatial information are represented: (1) time series data from resting-state functional MRI measuring brain activation; (2) network properties derived from structural connectivity data for different groups of subjects, which may help guide the detection of differentiation features. Through visual exploration, this integrated solution can help identify brain regions with highly correlated functional activations as well as their activation patterns. Visual detection of differentiation features can also potentially discover image-based phenotypic biomarkers for brain diseases.

Integrated Visualization of Human Brain Connectome Data.

Visualization plays a vital role in the analysis of multi-modal neuroimaging data. A major challenge in neuroimaging visualization is how to integrate structural, functional and connectivity data to form a comprehensive visual context for data exploration, quality control, and hypothesis discovery. We develop a new integrated visualization solution for brain imaging data by combining scientific and information visualization techniques within the context of the same anatomic structure. New surface texture techniques are developed to map non-spatial attributes onto the brain surfaces from MRI scans. Two types of non-spatial information are represented: (1) time-series data from resting-state functional MRI measuring brain activation; (2) network properties derived from structural connectivity data for different groups of subjects, which may help guide the detection of differentiation features. Through visual exploration, this integrated solution can help identify brain regions with highly correlated functional activations as well as their activation patterns. Visual detection of differentiation features can also potentially discover image based phenotypic biomarkers for brain diseases.

Feature decomposition architectures for neural networks: algorithms, error bounds, and applications.

In recent years, systems consisting of multiple modular neural networks have attracted substantial interest in the neural networks community because of various advantages they offer over a single large monolithic network. In this paper, we propose two basic feature decomposition models (namely, parallel model and tandem model) in which each of the neural network modules processes a disjoint subset of the input features. A novel feature decomposition algorithm is introduced to partition the input space into disjoint subsets solely based on the available training data. Under certain assumptions, the approximation error due to decomposition can be proved to be bounded by any desired small value over a compact set. Finally, the performance of feature decomposition networks is compared with that of a monolithic network in real world bench mark pattern recognition and modeling problems.

Identifying the neuroanatomical basis of cognitive impairment in Alzheimer's disease by correlation- and nonlinearity-aware sparse Bayesian learning.

Predicting cognitive performance of subjects from their magnetic resonance imaging (MRI) measures and identifying relevant imaging biomarkers are important research topics in the study of Alzheimer's disease. Traditionally, this task is performed by formulating a linear regression problem. Recently, it is found that using a linear sparse regression model can achieve better prediction accuracy. However, most existing studies only focus on the exploitation of sparsity of regression coefficients, ignoring useful structure information in regression coefficients. Also, these linear sparse models may not capture more complicated and possibly nonlinear relationships between cognitive performance and MRI measures. Motivated by these observations, in this work we build a sparse multivariate regression model for this task and propose an empirical sparse Bayesian learning algorithm. Different from existing sparse algorithms, the proposed algorithm models the response as a nonlinear function of the predictors by extending the predictor matrix with block structures. Further, it exploits not only inter-vector correlation among regression coefficient vectors, but also intra-block correlation in each regression coefficient vector. Experiments on the Alzheimer's Disease Neuroimaging Initiative database showed that the proposed algorithm not only achieved better prediction performance than state-of-the-art competitive methods, but also effectively identified biologically meaningful patterns.

Pattern Visualization of Human Connectome Data.

The human brain is a complex network with countless connected neurons, and can be described as a "connectome". Existing studies on analyzing human connectome data are primarily focused on characterizing the brain networks with a small number of easily computable measures that may be inadequate for revealing complex relationship between brain function and its structural substrate. To facilitate large-scale connectomic analysis, in this paper, we propose a powerful and flexible volume rendering scheme to effectively visualize and interactively explore thousands of network measures in the context of brain anatomy, and to aid pattern discovery. We demonstrate the effectiveness of the proposed scheme by applying it to a real connectome data set.

Hippocampal surface mapping of genetic risk factors in AD via sparse learning models.

Genetic mapping of hippocampal shape, an under-explored area, has strong potential as a neurodegeneration biomarker for AD and MCI. This study investigates the genetic effects of top candidate single nucleotide polymorphisms (SNPs) on hippocampal shape features as quantitative traits (QTs) in a large cohort. FS+LDDMM was used to segment hippocampal surfaces from MRI scans and shape features were extracted after surface registration. Elastic net (EN) and sparse canonical correlation analysis (SCCA) were proposed to examine SNP-QT associations, and compared with multiple regression (MR). Although similar in power, EN yielded substantially fewer predictors than MR. Detailed surface mapping of global and localized genetic effects were identified by MR and EN to reveal multi-SNP-single-QT relationships, and by SCCA to discover multi-SNP-multi-QT associations. Shape analysis identified stronger SNP-QT correlations than volume analysis. Sparse multivariate models have greater power to reveal complex SNP-QT relationships. Genetic analysis of quantitative shape features has considerable potential for enhancing mechanistic understanding of complex disorders like AD.

Facial Image Classification of Mouse Embryos for the Animal Model Study of Fetal Alcohol Syndrome.

Fetal Alcohol Syndrome (FAS) is a developmental disorder caused by maternal drinking during pregnancy. Computerize imaging techniques have been applied to study human facial dysmorphology associated with FAS. This paper describes a new facial image analysis method based on a multi-angle image classification technique using micro-video images of mouse embryo. Images taken from several different angles are analyzed separately, and the results are combined for classifications that separate embryos with and without alcohol exposures. Analysis results from animal models provide critical references for the understanding of FAS and potential therapy solutions for human patients.

Zero-crossing iso-surfaces in volume datasets.

This paper describes a new surface extraction method for volume datasets by applying iso-surface extraction techniques to the zero-crossing edges in a volumetric domain. A volume dataset is first filtered using a Laplacian of Gaussian filter to generate a zero-crossing field. A marching cube process will then be able to extract the entire zero-crossing surface that may be viewed selectively based on various intensity ranges and gradient scales. This new technique provides a more efficient surface navigation and extraction mechanism, as well as more accurate surface details, than the traditional iso-surface techniques.

Voxx: a PC-based, near real-time volume rendering system for biological microscopy.

Confocal and two-photon fluorescence microscopy have advanced the exploration of complex, three-dimensional biological structures at submicron resolution. We have developed a voxel-based three-dimensional (3-D) imaging program (Voxx) capable of near real-time rendering that runs on inexpensive personal computers. This low-cost interactive 3-D imaging system provides a powerful tool for analyzing complex structures in cells and tissues and encourages a more thorough exploration of complex biological image data.