Search

As photoacoustic imaging (PAI) technology matures, computational modeling will increasingly represent a critical tool for facilitating clinical translation through predictive simulation of real-world performance under a wide range of device and biological conditions. While modeling currently offers a rapid, inexpensive tool for device development and prediction of fundamental image quality metrics (e.g., spatial resolution and contrast ratio), rigorous verification and validation will be required of models used to provide regulatory-grade data that effectively complements and/or replaces in vivo testing. To address methods for establishing model credibility, we developed an integrated computational model of PAI by coupling a previously developed three-dimensional Monte Carlo model of tissue light transport with a two-dimensional (2D) acoustic wave propagation model implemented in the well-known k-Wave toolbox. We then evaluated ability of the model to predict basic image quality metrics by applying standardized verification and validation principles for computational models. The model was verified against published simulation data and validated against phantom experiments using a custom PAI system. Furthermore, we used the model to conduct a parametric study of optical and acoustic design parameters. Results suggest that computationally economical 2D acoustic models can adequately predict spatial resolution, but metrics such as signal-to-noise ratio and penetration depth were difficult to replicate due to challenges in modeling strong clutter observed in experimental images. Parametric studies provided quantitative insight into complex relationships between transducer characteristics and image quality as well as optimal selection of optical beam geometry to ensure adequate image uniformity. Multidomain PAI simulation tools provide high-quality tools to aid device development and prediction of real-world performance, but further work is needed to improve model fidelity, especially in reproducing image noise and clutter.

Accurate assessment of burn severity is critical for wound care and the course of treatment. Delays in classification translate to delays in burn management, increasing the risk of scarring and infection. To this end, numerous imaging techniques have been used to examine tissue properties to infer burn severity. Spatial frequency-domain imaging (SFDI) has also been used to characterize burns based on the relationships between histologic observations and changes in tissue properties. Recently, machine learning has been used to classify burns by combining optical features from multispectral or hyperspectral imaging. Rather than employ models of light propagation to deduce tissue optical properties, we investigated the feasibility of using SFDI reflectance data at multiple spatial frequencies, with a support vector machine (SVM) classifier, to predict severity in a porcine model of graded burns. Calibrated reflectance images were collected using SFDI at eight wavelengths (471 to 851 nm) and five spatial frequencies (0 to 0.2 mm−1). Three models were built from subsets of this initial dataset. The first subset included data taken at all wavelengths with the planar (0 mm−1) spatial frequency, the second comprised data at all wavelengths and spatial frequencies, and the third used all collected data at values relative to unburned tissue. These data subsets were used to train and test cubic SVM models, and compared against burn status 28 days after injury. Model accuracy was established through leave-one-out cross-validation testing. The model based on images obtained at all wavelengths and spatial frequencies predicted burn severity at 24 h with 92.5% accuracy. The model composed of all values relative to unburned skin was 94.4% accurate. By comparison, the model that employed only planar illumination was 88.8% accurate. This investigation suggests that the combination of SFDI with machine learning has potential for accurately predicting burn severity.

We study a problem scenario of super-resolution (SR) algorithms in the context of whole slide imaging (WSI), a popular imaging modality in digital pathology. Instead of just one pair of high- and low-resolution images, which is typically the setup in which SR algorithms are designed, we are given multiple intermediate resolutions of the same image as well. The question remains how to best utilize such data to make the transformation learning problem inherent to SR more tractable and address the unique challenges that arises in this biomedical application. We propose a recurrent convolutional neural network model, to generate SR images from such multi-resolution WSI datasets. Specifically, we show that having such intermediate resolutions is highly effective in making the learning problem easily trainable and address large resolution difference in the low and high-resolution images common in WSI, even without the availability of a large size training data. Experimental results show state-of-the-art performance on three WSI histopathology cancer datasets, across a number of metrics.

Digital video now plays an important role in supporting more profitable online patient training and counseling, and integration of patient training videos from multiple competitive organizations in the health care network will result in better offerings for patients. However, privacy concerns often prevent multiple competitive organizations from sharing and integrating their patient training videos. In addition, patients with infectious or chronic diseases may not want the online patient training organizations to identify who they are or even which video clips they are interested in. Thus, there is an urgent need to develop more effective techniques to protect both video content privacy and access privacy . In this paper, we have developed a new approach to construct a distributed Hippocratic video database system for supporting more profitable online patient training and counseling. First, a new database modeling approach is developed to support concept-oriented video database organization and assign a degree of privacy of the video content for each database level automatically. Second, a new algorithm is developed to protect the video content privacy at the level of individual video clip by filtering out the privacy-sensitive human objects automatically. In order to integrate the patient training videos from multiple competitive organizations for constructing a centralized video database indexing structure, a privacy-preserving video sharing scheme is developed to support privacy-preserving distributed classifier training and prevent the statistical inferences from the videos that are shared for cross-validation of video classifiers. Our experiments on large-scale video databases have also provided very convincing results.

During a meeting of the SYSGENET working group 'Bioinformatics', currently available software tools and databases for systems genetics in mice were reviewed and the needs for future developments discussed. The group evaluated interoperability and performed initial feasibility studies. To aid future compatibility of software and exchange of already developed software modules, a strong recommendation was made by the group to integrate HAPPY and R/qtl analysis toolboxes, GeneNetwork and XGAP database platforms, and TIQS and xQTL processing platforms. R should be used as the principal computer language for QTL data analysis in all platforms and a 'cloud' should be used for software dissemination to the community. Furthermore, the working group recommended that all data models and software source code should be made visible in public repositories to allow a coordinated effort on the use of common data structures and file formats.

Survival analysis is a popular branch of statistics. At present, many algorithms (like traditional multi-tasking learning model) cannot be applied well in practice because of censored data. Although using some model (like parametric regression model) can avoid it, they need strict assumptions. This undermines the very nature of things, which is very detrimental to the study of practical problems. The method proposed in this paper can apply well to the censored data, but does not need to make any additional assumptions about the original problem. It can be said that it breaks through the above two kinds of major limitations. The algorithm is a kind of inductive transfer learning method, which can fully obtain the information in the censored data, using domain-specific information implicit in each feature to enhance the generalization capability of the model. We also used two common performance metrics as criteria to judge the predictive performance differences between the models in this article and those of other mainstream models. The results show that the model proposed in this paper is 10 ∼ 15 percent higher than other mainstream models, which proves that our multi-task learning model has a great advantage in the survival analysis of cancer genes.

Objective : Multi-modal brain functional connectivity (FC) data have shown great potential for providing insights into individual variations in behavioral and cognitive traits. The joint learning of multi-modal data can utilize intrinsic association, and thus can boost learning performance. Although several multi-task based learning models have already been proposed by viewing feature learning on each modality as one task, most of them ignore the structural information inherent across the modalities, which may play an important role in extracting discriminative features. Methods : In this paper, we propose a new manifold regularized multi-task learning model by simultaneously considering between-subject and between-modality relationships. Specifically, the $l_{2,1}$ -norm (i.e., group-sparsity) regularizer is enforced to jointly select a few common features across different modalities. A novelly designed manifold regularizer is further imposed as a crucial underpinning to preserve the structural information both within and between modalities. Such designed regularizers will make our model more adaptive to realistic neuroimaging data, which are usually of small sample size but high dimensional features. Results : Our model is validated on the Philadelphia Neurodevelopmental Cohort dataset, where our modalities are regarded as two types of functional MRI (fMRI) data collected under two paradigms. We conduct experimental studies on fMRI-based FC network data in two task conditions for intelligence quotient (IQ) prediction. The results show that our proposed model can not only achieve improved prediction performance, but also yield a set of IQ-relevant biomarkers. Conclusion and Significance : This paper develops a new multi-task learning model, enabling the discovery of significant biomarkers that may account for a proportion of the variance in human intelligence.

This paper develops an effective distributed strategy for the solution of constrained multiagent stochastic optimization problems with coupled parameters across the agents. In this formulation, each agent is influenced by only a subset of the entries of a global parameter vector or model, and is subject to convex constraints that are only known locally. Problems of this type arise in several applications, most notably in disease propagation models, minimum-cost flow problems, distributed control formulations, and distributed power system monitoring. This paper focuses on stochastic settings, where a stochastic risk function is associated with each agent and the objective is to seek the minimizer of the aggregate sum of all risks subject to a set of constraints. Agents are not aware of the statistical distribution of the data and, therefore, can only rely on stochastic approximations in their learning strategies. We derive an effective distributed learning strategy that is able to track drifts in the underlying parameter model. A detailed performance and stability analysis is carried out showing that the resulting coupled diffusion strategy converges at a linear rate to an $O(\mu)$ neighborhood of the true penalized optimizer.

This paper proposes a novel approach for online, individualized gait analysis, based on an adaptive periodic model of any gait signal. The proposed method learns a model of the gait cycle during online measurement, using a continuous representation that can adapt to inter- and intra-personal variability by creating an individualized model. Once the algorithm has converged to the input signal, key gait events can be identified based on the estimated gait phase and amplitude. The approach is implemented and tested on retirement home resident 6 min walk (6MW) data using wearable accelerometers at the ankle. The proposed approach converges within approximately four gait cycles and achieves 3% error in detecting initial swing events. 1 1 An early version of this work was presented in [1] . A more extensive description of related work and an extended method, including optimization of learning rates, were added to this paper. Further, this paper applies and evaluates the method to a new and much larger gait dataset taken from older adults who each have a variety of medical conditions. Therefore, the experimental protocol was also updated and the results are entirely novel.

The growing popularity of big data analysis and cloud computing has created new big data management standards. Sometimes, programmers may interact with a number of heterogeneous data stores depending on the information they are responsible for: SQL and NoSQL data stores. Interacting with heterogeneous data models via numerous APIs and query languages imposes challenging tasks on multi-data processing developers. Indeed, complex queries concerning homogenous data structures cannot currently be performed in a declarative manner when found in single data storage applications and therefore require additional development efforts. Many models were presented in order to address complex queries Via multistore applications. Some of these models implemented a complex unified and fast model, while others’ efficiency is not good enough to solve this type of complex database queries. This paper provides an automated, fast and easy unified architecture to solve simple and complex SQL and NoSQL queries over heterogeneous data stores (CQNS). This proposed framework can be used in cloud environments or for any big data application to automatically help developers to manage basic and complicated database queries. CQNS consists of three layers: matching selector layer, processing layer, and query execution layer. The matching selector layer is the heart of this architecture in which five of the user queries are examined if they are matched with another five queries stored in a single engine stored in the architecture library. This is achieved through a proposed algorithm that directs the query to the right SQL or NoSQL database engine. Furthermore, CQNS deal with many NoSQL Databases like MongoDB, Cassandra, Riak, CouchDB, and NOE4J databases. This paper presents a spark framework that can handle both SQL and NoSQL Databases. Four scenarios’ benchmarks datasets are used to evaluate the proposed CQNS for querying different NoSQL Databases in terms of optimization process performance and query execution time. The results show that, the CQNS achieves best latency and throughput in less time among the compared systems.

Prediction of seizure before they occur is vital for bringing normalcy to the lives of patients. Researchers employed machine learning methods using hand-crafted features for seizure prediction. However, ML methods are too complicated to select the best ML model or best features. Deep Learning methods are beneficial in the sense of automatic feature extraction. One of the roadblocks for accurate seizure prediction is scarcity of epileptic seizure data. This paper addresses this problem by proposing a deep convolutional generative adversarial network to generate synthetic EEG samples. We use two methods to validate synthesized data namely, one-class SVM and a new proposal which we refer to as convolutional epileptic seizure predictor (CESP). Another objective of our study is to evaluate performance of well-known deep learning models (e.g., VGG16, VGG19, ResNet50, and Inceptionv3) by training models on augmented data using transfer learning with average time of 10 min between true prediction and seizure onset. Our results show that CESP model achieves sensitivity of 78.11% and 88.21%, and FPR of 0.27/h and 0.14/h for training on synthesized and testing on real Epilepsyecosystem and CHB-MIT datasets, respectively. Effective results of CESP trained on synthesized data shows that synthetic data acquired the correlation between features and labels very well. We also show that employment of idea of transfer learning and data augmentation in patient-specific manner provides highest accuracy with sensitivity of 90.03% and 0.03 FPR/h which was achieved using Inceptionv3, and that augmenting data with samples generated from DCGAN increased prediction results of our CESP model and Inceptionv3 by 4-5% as compared to state-of-the-art traditional augmentation techniques. Finally, we note that prediction results of CESP achieved by using augmented data are better than chance level for both datasets., 10 pages, 5 figures, 6 Tables, Journal paper

Background and objective: Cancer, as the most challenging part in the human disease history, has always been one of the main threats to human life and health. The high mortality of cancer is largely due to the complexity of cancer and the significant differences in clinical outcomes. Therefore, it will be significant to improve accuracy of cancer survival prediction, which has become one of the main fields of cancer research. Many calculation models for cancer survival prediction have been proposed at present, but most of them generate prediction models only by using single genomic data or clinical data. Multiple genomic data and clinical data have not been integrated yet to take a comprehensive consideration of cancers and predict their survival. Method: In order to effectively integrate multiple genomic data (including genetic expression, copy number alteration, DNA methylation and exon expression) and clinical data and apply them to predictive studies on cancer survival, similar network fusion algorithm (SNF) was proposed in this paper to integrate multiple genomic data and clinical data so as to generate sample similarity matrix, min-redundancy and max-relevance algorithm (mRMR) was used to conduct feature selection of multiple genomic data and clinical data of cancer samples and generate sample feature matrix, and finally two matrixes were used for semi-supervised training through graph convolutional network (GCN) so as to obtain a cancer survival prediction method integrating multiple genomic data and clinical data based on graph convolutional network (GCGCN). Result: Performance indexes of GCGCN model indicate that both multiple genomic data and clinical data play significant roles in the accurate survival time prediction of cancer patients. It is compared with existing survival prediction methods, and results show that cancer survival prediction method GCGCN which integrates multiple genomic data and clinical data has obviously superior prediction effect than existing survival prediction methods. Conclusion: All study results in this paper have verified effectiveness and superiority of GCGCN in the aspect of cancer survival prediction.

With the era of big data, the utilization of machine learning algorithms in radiation oncology is rapidly growing with applications including: treatment response modeling, treatment planning, contouring, organ segmentation, image-guidance, motion tracking, quality assurance, and more. Despite this interest, practical clinical implementation of machine learning as part of the day-to-day clinical operations is still lagging. The aim of this white paper is to further promote progress in this new field of machine learning in radiation oncology by highlighting its untapped advantages and potentials for clinical advancement, while also presenting current challenges and open questions for future research. The targeted audience of this paper includes newcomers as well as practitioners in the field of medical physics/radiation oncology. The paper also provides general recommendations to avoid common pitfalls when applying these powerful data analytic tools to medical physics and radiation oncology problems and suggests some guidelines for transparent and informative reporting of machine learning results.

In this paper, we study Copy Number Variation (CNV) data. The underlying process generating CNV segments is generally assumed to be memory-less, giving rise to an exponential distribution of segment lengths. In this paper, we provide evidence from cancer patient data, which suggests that this generative model is too simplistic, and that segment lengths follow a power-law distribution instead. We conjecture a simple preferential attachment generative model that provides the basis for the observed power-law distribution. We then show how an existing statistical method for detecting cancer driver genes can be improved by incorporating the power-law distribution in the null model.

Unsupervised domain adaptation (UDA) is to learn classification models that make predictions for unlabeled data on a target domain, given labeled data on a source domain whose distribution diverges from the target one. Mainstream UDA methods strive to learn domain-aligned features such that classifiers trained on the source features can be readily applied to the target ones. Although impressive results have been achieved, these methods have a potential risk of damaging the intrinsic data structures of target discrimination, raising an issue of generalization particularly for UDA tasks in an inductive setting. To address this issue, we are motivated by a UDA assumption of structural similarity across domains, and propose to directly uncover the intrinsic target discrimination via constrained clustering, where we constrain the clustering solutions using structural source regularization that hinges on the very same assumption. Technically, we propose a hybrid model of Structurally Regularized Deep Clustering, which integrates the regularized discriminative clustering of target data with a generative one, and we thus term our method as H-SRDC. Our hybrid model is based on a deep clustering framework that minimizes the Kullback-Leibler divergence between the distribution of network prediction and an auxiliary one, where we impose structural regularization by learning domain-shared classifier and cluster centroids. By enriching the structural similarity assumption, we are able to extend H-SRDC for a pixel-level UDA task of semantic segmentation. We conduct extensive experiments on seven UDA benchmarks of image classification and semantic segmentation. With no explicit feature alignment, our proposed H-SRDC outperforms all the existing methods under both the inductive and transductive settings. We make our implementation codes publicly available at https://github.com/huitangtang/H-SRDC., Journal extension of our preliminary CVPR conference paper, under review, 16 pages, 8 figures, 9 tables

As pairwise ranking becomes broadly employed for elections, sports competitions, recommendations, and so on, attackers have strong motivation and incentives to manipulate the ranking list. They could inject malicious comparisons into the training data to fool the victim. Such a technique is called poisoning attack in regression and classification tasks. In this paper, to the best of our knowledge, we initiate the first systematic investigation of data poisoning attacks on pairwise ranking algorithms, which can be formalized as the dynamic and static games between the ranker and the attacker and can be modeled as certain kinds of integer programming problems. To break the computational hurdle of the underlying integer programming problems, we reformulate them into the distributionally robust optimization (DRO) problems, which are computationally tractable. Based on such DRO formulations, we propose two efficient poisoning attack algorithms and establish the associated theoretical guarantees. The effectiveness of the suggested poisoning attack strategies is demonstrated by a series of toy simulations and several real data experiments. These experimental results show that the proposed methods can significantly reduce the performance of the ranker in the sense that the correlation between the true ranking list and the aggregated results can be decreased dramatically., 31 pages

One-class learning is the classic problem of fitting a model to the data for which annotations are available only for a single class. In this paper, we explore novel objectives for one-class learning, which we collectively refer to as Generalized One-class Discriminative Subspaces (GODS). Our key idea is to learn a pair of complementary classifiers to flexibly bound the one-class data distribution, where the data belongs to the positive half-space of one of the classifiers in the complementary pair and to the negative half-space of the other. To avoid redundancy while allowing non-linearity in the classifier decision surfaces, we propose to design each classifier as an orthonormal frame and seek to learn these frames via jointly optimizing for two conflicting objectives, namely: i) to minimize the distance between the two frames, and ii) to maximize the margin between the frames and the data. The learned orthonormal frames will thus characterize a piecewise linear decision surface that allows for efficient inference, while our objectives seek to bound the data within a minimal volume that maximizes the decision margin, thereby robustly capturing the data distribution. We explore several variants of our formulation under different constraints on the constituent classifiers, including kernelized feature maps. We demonstrate the empirical benefits of our approach via experiments on data from several applications in computer vision, such as anomaly detection in video sequences, human poses, and human activities. We also explore the generality and effectiveness of GODS for non-vision tasks via experiments on several UCI datasets, demonstrating state-of-the-art results., Accepted at Trans. PAMI. arXiv admin note: text overlap with arXiv:1908.05884

With substantial amount of time, resources and human (team) efforts invested to explore and develop successful deep neural networks (DNN), there emerges an urgent need to protect these inventions from being illegally copied, redistributed, or abused without respecting the intellectual properties of legitimate owners. Following recent progresses along this line, we investigate a number of watermark-based DNN ownership verification methods in the face of ambiguity attacks, which aim to cast doubts on the ownership verification by forging counterfeit watermarks. It is shown that ambiguity attacks pose serious threats to existing DNN watermarking methods. As remedies to the above-mentioned loophole, this paper proposes novel passport-based DNN ownership verification schemes which are both robust to network modifications and resilient to ambiguity attacks. The gist of embedding digital passports is to design and train DNN models in a way such that, the DNN inference performance of an original task will be significantly deteriorated due to forged passports. In other words, genuine passports are not only verified by looking for the predefined signatures, but also reasserted by the unyielding DNN model inference performances. Extensive experimental results justify the effectiveness of the proposed passport-based DNN ownership verification schemes. Code is available at https://github.com/kamwoh/DeepIPR.

Scientists conducting microarray and other experiments use circular Venn and Euler diagrams to analyze and illustrate their results. As one solution to this problem, this paper introduces a statistical model for fitting area-proportional Venn and Euler diagrams to observed data. The statistical model outlined in this paper includes a statistical loss function and a minimization procedure that enables formal estimation of the Venn/Euler area-proportional model for the first time. A significance test of the null hypothesis is computed for the solution. Residuals from the model are available for inspection. As a result, this algorithm can be used for both exploration and inference on real data sets. A Java program implementing this algorithm is available under the Mozilla Public License. An R function venneuler () is available as a package in CRAN and a plugin is available in Cytoscape.

Research papers and data products are key outcomes of the science enterprise. Governmental, nongovernmental, and private foundation sponsors of research are increasingly recognizing the value of research data. As a result, most funders now require that sufficiently detailed data management plans be submitted as part of a research proposal. A data management plan (DMP) is a document that describes how you will treat your data during a project and what happens with the data after the project ends. Such plans typically cover all or portions of the data life cycle—from data discovery, collection, and organization (e.g., spreadsheets, databases), through quality assurance/quality control, documentation (e.g., data types, laboratory methods) and use of the data, to data preservation and sharing with others (e.g., data policies and dissemination approaches). Fig 1 illustrates the relationship between hypothetical research and data life cycles and highlights the links to the rules presented in this paper. The DMP undergoes peer review and is used in part to evaluate a project’s merit. Plans also document the data management activities associated with funded projects and may be revisited during performance reviews. Open in a separate window Fig 1 Relationship of the research life cycle (A) to the data life cycle (B); note: highlighted circles refer to the rules that are most closely linked to the steps of the data life cycle. As part of the research life cycle (A), many researchers (1) test ideas and hypotheses by (2) acquiring data that are (3) incorporated into various analyses and visualizations, leading to interpretations that are then (4) published in the literature and disseminated via other mechanisms (e.g., conference presentations, blogs, tweets), and that often lead back to (1) new ideas and hypotheses. During the data life cycle (B), researchers typically (1) develop a plan for how data will be managed during and after the project; (2) discover and acquire existing data and (3) collect and organize new data; (4) assure the quality of the data; (5) describe the data (i.e., ascribe metadata); (6) use the data in analyses, models, visualizations, etc.; and (7) preserve and (8) share the data with others (e.g., researchers, students, decision makers), possibly leading to new ideas and hypotheses.

Purely data driven approaches for machine learning present difficulties when data is scarce relative to the complexity of the model or when the model is forced to extrapolate. On the other hand, purely mechanistic approaches need to identify and specify all the interactions in the problem at hand (which may not be feasible) and still leave the issue of how to parameterize the system. In this paper, we present a hybrid approach using Gaussian processes and differential equations to combine data driven modelling with a physical model of the system. We show how different, physically-inspired, kernel functions can be developed through sensible, simple, mechanistic assumptions about the underlying system. The versatility of our approach is illustrated with three case studies from motion capture, computational biology and geostatistics., 20 pages, 2 figures. Extended technical report of the Conference Paper "Latent force models" in D. van Dyk and M. Welling (eds) Proceedings of the Twelfth International Workshop on Artificial Intelligence and Statistics, JMLR W&CP 5, Clearwater Beach, FL, pp 9--16

The past two decades have witnessed significant advances in high-throughput “omics” technologies such as genomics, proteomics, metabolomics, transcriptomics and radiomics. These technologies have enabled simultaneous measurement of the expression levels of tens of thousands of features from individual patient samples and have generated enormous amounts of data that require analysis and interpretation. One specific area of interest has been in studying the relationship between these features and patient outcomes, such as overall and recurrence-free survival, with the goal of developing a predictive “omics” profile. Large-scale studies often suffer from the presence of a large fraction of censored observations and potential time-varying effects of features, and methods for handling them have been lacking. In this paper, we propose supervised methods for feature selection and survival prediction that simultaneously deal with both issues. Our approach utilizes continuum power regression (CPR) - a framework that includes a variety of regression methods - in conjunction with the parametric or semi-parametric accelerated failure time (AFT) model. Both CPR and AFT fall within the linear models framework and, unlike black-box models, the proposed prognostic index has a simple yet useful interpretation. We demonstrate the utility of our methods using simulated and publicly available cancer genomics data.

For most problems in science and engineering we can obtain data sets that describe the observed system from various perspectives and record the behavior of its individual components. Heterogeneous data sets can be collectively mined by data fusion. Fusion can focus on a specific target relation and exploit directly associated data together with contextual data and data about system's constraints. In the paper we describe a data fusion approach with penalized matrix tri-factorization (DFMF) that simultaneously factorizes data matrices to reveal hidden associations. The approach can directly consider any data that can be expressed in a matrix, including those from feature-based representations, ontologies, associations and networks. We demonstrate the utility of DFMF for gene function prediction task with eleven different data sources and for prediction of pharmacologic actions by fusing six data sources. Our data fusion algorithm compares favorably to alternative data integration approaches and achieves higher accuracy than can be obtained from any single data source alone., Comment: Short preprint, 13 pages, 3 Figures, 3 Tables. Full paper in 10.1109/TPAMI.2014.2343973

This special section contains the second series of journal articles made from a selection of papers presented at the Ninth International Conference on Computational Methods in Systems Biology, CMSB 2011. CMSB is an annual series of conferences, initiated in 2003, on the design of computational methods for modeling and analyzing biological systems, networks, data, and on their applications to study cases. This issue contains the second part which is composed of three papers.“The Propagation Approach for Computing Biochemical Reaction Networks” by Thomas A. Henzinger and Maria Mateescu revisits the chemical master equation, the rate equation and a combination of both, with the concepts of propagation models and propagation data type for abstracting from the further implementation choices made in simulators.“Curvature Analysis of Cardiac Excitation Wavefronts” by Abhishek Murthy, Ezio Bartocci, Flavio H. Fenton, James Glimm, Richard A. Gray, Elizabeth M. Cherry, Scott A. Smolka, and Radu Grosu, which describes a parallel curvature analysis algorithm of cardiac excitation wavefronts.The last paper, “Multiscale Modeling and Analysis of Planar Cell Polarity in the Drosophila Wing,” by Qian Gao, David Gilbert, Monika Heiner, Fei Liu, Daniele Maccagnola, and David Tree proposes the use of Hierarchically Colored Petri Nets to develop models of tissues at different spatial scales.

High-risk prediction of cardiovascular disease is of great significance and impendency in medical fields with the increasing phenomenon of sub-health these years. Most existing pathological methods for the prognosis prediction are either costly or prone to misjudgement. Therefore, plenty of automated models based on machine learning have been proposed to predict the onset of cardiovascular disease with the premorbid information of patients extracted from their historical Electronic Health Records (EHRs). However, it is a tough job to select proper features from longitudinal and heterogeneous EHRs, and also a great challenge to obtain accurate and robust representations for patients. In this paper, we propose an entirely end-to-end model called DeepRisk based on attention mechanism and deep neural networks, which can not only learn high-quality features automatically from EHRs, but also efficiently integrate heterogeneous and time-ordered medical data, and finally predict patients' risk of cardiovascular diseases. Experiments are carried out on a real medical dataset and results show that DeepRisk can significantly improve the high-risk prediction accuracy for cardiovascular disease compared with state-of-the-art approaches.

Object parsing and segmentation from point clouds are challenging tasks because the relevant data is available only as thin structures along object boundaries or other features, and is corrupted by large amounts of noise. To handle this kind of data, flexible shape models are desired that can accurately follow the object boundaries. Popular models such as Active Shape and Active Appearance models lack the necessary flexibility for this task, while recent approaches such as the Recursive Compositional Models make model simplifications in order to obtain computational guarantees. This paper investigates a hierarchical Bayesian model of shape and appearance in a generative setting. The input data is explained by an object parsing layer, which is a deformation of a hidden PCA shape model with Gaussian prior. The paper also introduces a novel efficient inference algorithm that uses informed data-driven proposals to initialize local searches for the hidden variables. Applied to the problem of object parsing from structured point clouds such as edge detection images, the proposed approach obtains state of the art parsing errors on two standard datasets without using any intensity information., 13 pages, 16 figures

Different biological labs conduct similar experiments on same diseases. It is highly desirable to have a better model based on more experimental results than that on a single result. In this paper, we propose a method for integrating microarray data from multiple sources for building classification models. To test the method, we use three real world microarray data sets from different labs with different experimental devices and environments. Although microarray data is well known for its inconsistencies across labs, we demonstrate that it is possible to build consistent models using data sets from multiple labs. We report our method, experimental results and observations in the paper. Refereed/Peer-reviewed

As data sources become larger and more complex, the ability to effectively explore and analyze patterns amongst varying sources becomes a critical bottleneck in analytic reasoning. Incoming data contains multiple variables, high signal to noise ratio, and a degree of uncertainty, all of which hinder exploration, hypothesis eneration/exploration, and decision making. To facilitate the exploration of such data, advanced tool sets are needed that allow the user to interact with their data in a visual environment that provides direct analytic capability for finding data aberrations or hotspots. In this paper, we present a suite of tools designed to facilitate the exploration of spatiotemporal datasets. Our system allows users to search for hotspots in both space and time, combining linked views and interactive filtering to provide users with contextual information about their data and allow the user to develop and explore their hypotheses. Statistical data models and alert detection algorithms are provided to help draw user attention to critical areas. Demographic filtering can then be further applied as hypotheses generated become fine tuned. This paper demonstrates the use of such tools on multiple geo-spatiotemporal datasets.

Ultra-high definition (UHD) 360 videos encoded in fine quality are typically too large to stream in its entirety over bandwidth (BW)-constrained networks. One popular approach is to interactively extract and send a spatial sub-region corresponding to a viewer's current field-of-view (FoV) in a head-mounted display (HMD) for more BW-efficient streaming. Due to the non-negligible round-trip-time (RTT) delay between server and client, accurate head movement prediction foretelling a viewer's future FoVs is essential. In this paper, we cast the head movement prediction task as a sparse directed graph learning problem: three sources of relevant information-collected viewers' head movement traces, a 360 image saliency map, and a biological human head model-are distilled into a view transition Markov model. Specifically, we formulate a constrained maximum a posteriori (MAP) problem with likelihood and prior terms defined using the three information sources. We solve the MAP problem alternately using a hybrid iterative reweighted least square (IRLS) and Frank-Wolfe (FW) optimization strategy. In each FW iteration, a linear program (LP) is solved, whose runtime is reduced thanks to warm start initialization. Having estimated a Markov model from data, we employ it to optimize a tile-based 360 video streaming system. Extensive experiments show that our head movement prediction scheme noticeably outperformed existing proposals, and our optimized tile-based streaming scheme outperformed competitors in rate-distortion performance.

The performance of a convolutional neural network (CNN) based face recognition model largely relies on the richness of labeled training data. However, it is expensive to collect a training set with large variations of a face identity under different poses and illumination changes, so the diversity of within-class face images becomes a critical issue in practice. In this paper, we propose a 3D model-assisted domain-transferred face augmentation network (DotFAN) that can generate a series of variants of an input face based on the knowledge distilled from existing rich face datasets of other domains. Extending from StarGAN's architecture, DotFAN integrates with two additional subnetworks, i.e., face expert model (FEM) and face shape regressor (FSR), for latent facial code control. While FSR aims to extract face attributes, FEM is designed to capture a face identity. With their aid, DotFAN can separately learn facial feature codes and effectively generate face images of various facial attributes while keeping the identity of augmented faces unaltered. Experiments show that DotFAN is beneficial for augmenting small face datasets to improve their within-class diversity so that a better face recognition model can be learned from the augmented dataset.

Data simulation is an important research tool to evaluate algorithms. Two types of methods are currently used to simulate medical ultrasound data: those based on acoustic models and those based on convolution models. The simulation of ultrasound data sequences is very time-consuming. In addition, many applications require accounting for the out-of-plane motion induced by the 3-D displacement of scatterers. The purpose of this paper is to propose a model adapted to a fast simulation of ultrasonic data sequences with 3-D moving scatterers. Our approach is based on the convolution model. The scatterers are moved in a 3-D continuous medium between each pair of images and then projected onto the imaging plane before being convolved. This paper discusses the practical implementation of the convolution that can be performed directly or after a grid approximation. The grid approximation convolution is obviously faster than the direct convolution but generates errors resulting from the approximation to the grid's nodes. We provide the analytical expression of these errors and then define 2 intensity-based criteria to quantify them as a function of the spatial sampling. The simulation of an image requires less than 2 s with oversampling, thus reducing these errors. The simulation model is validated with first- and second-order statistics. The positions of the scatterers at each imaging time can be provided by a displacement model. An example applied to flow imaging is proposed. Several cases are used to show that this displacement model provides realistic data. It is validated with speckle tracking, a well-known motion estimator in ultrasound imaging.

Output-time prediction is a critical task to a wafer fab. To further enhance the accuracy of wafer-lot output-time prediction, the concept of input classification is applied to Chen's fuzzy backpropagation network (FBPN) approach in this paper by preclassifying wafer lots with the fuzzy c-means (FCM) classifier before predicting the output times. In this way, similar wafer lots are clustered in the same category. The data of wafer lots of different categories are then learned with different FBPNs but with the same topology. After learning, these FBPNs form an FBPN ensemble that can be applied in predicting the output time of a new lot. The output of the FBPN ensemble determines the cycle/output time forecast and is obtained by aggregating the outputs of the component FBPNs. Production simulation is applied in this paper to generate test data. According to experimental results, the prediction accuracy of the hybrid FCM-FBPN approach was significantly better than those of many existing approaches.

One of the main objectives of survival analysis is to predict the failure time that is usually considered as a continuous variable. In longitudinal studies, the data are often collected at every certain period of time, for example, monthly, quarterly, or yearly. Such data require appropriate techniques to handle the discrete time values that often have incomplete information about the failure occurrence-so-called "censored cases." Tree-based models are common, assumption-free methods of survival prediction. In this paper, the author proposes three recursive partitioning techniques able to cope with discrete-time censored survival data, which, in contrast to already-existing models limited to univariate trees, allow splits to have a form of any hyperplane. The performance of proposed methods, expressed as a mean absolute error, was examined on the basis of both synthetic and real data sets available in the literature and compared with existing tree-based models. To demonstrate the applicability of the methods in identifying subgroups of patients with a similar survival experience and to assess the influence of covariates on the risk of failure, a Veteran's Administration lung cancer data set was used. The results confirm proposed models to be good prediction tools for discrete-time survival data.

Current deep supervised learning methods typically require large amounts of labeled data for training. Since there is a significant cost associated with clinical data acquisition and labeling, medical datasets used for training these models are relatively small in size. In this paper, we aim to alleviate this limitation by proposing a variational generative model along with an effective data augmentation approach that utilizes the generative model to synthesize data. In our approach, the model learns the probability distribution of image data conditioned on a latent variable and the corresponding labels. The trained model can then be used to synthesize new images for data augmentation. We demonstrate the effectiveness of the approach on two independent clinical datasets consisting of ultrasound images of the spine and magnetic resonance images of the brain. For the spine dataset, a baseline and a residual model achieve an accuracy of 85% and 92%, respectively, using our method compared to 78% and 83% using a conventional training approach for image classification task. For the brain dataset, a baseline and a U-net network achieve an accuracy of 84% and 88%, respectively, in Dice coefficient in tumor segmentation compared to 80% and 83% for the convention training approach.

In many embedded systems, we face the problem of correlating signals characterising device operation (e.g., performance parameters, anomalies) with events describing internal device activities. This leads to the investigation of two types of data: time series, representing signal periodic samples in a background of noise, and sporadic event logs. The correlation process must take into account clock inconsistencies between the data acquisition and monitored devices, which provide time series signals and event logs, respectively. The idea of the presented solution is to classify event logs based on the introduced similarity metric and deriving their distribution in time. The identified event log sequences are matched with time intervals corresponding to specified sample patterns (objects) in the registered signal time series. The matching (correlation) process involves iterative time offset adjustment. The paper presents original algorithms to investigate correlation problems using the object-oriented data models corresponding to two monitoring sources. The effectiveness of this approach has been verified in power consumption analysis using real data collected from the developed Holter device. It is quite universal and can be easily adapted to other device optimisation problems.

Point cloud registration is a key step in the reconstruction of 3D data models. The traditional ICP registration algorithm depends on the initial position of the point cloud. Otherwise, it may get trapped into local optima. In addition, the registration method based on the feature learning of PointNet cannot directly or effectively extract local features. To solve these two problems, this paper proposes SAP-Net, inspired by CorsNet and PointNet++, as an optimized CorsNet. To be more specific, SAP-Net firstly uses the set abstraction layer in PointNet++ as the feature extraction layer and then combines the global features with the initial template point cloud. Finally, PointNet is used as the transform prediction layer to obtain the six parameters required for point cloud registration directly, namely the rotation matrix and the translation vector. Experiments on the ModelNet40 dataset and real data show that SAP-Net not only outperforms ICP and CorsNet on both seen and unseen categories of the point cloud but also has stronger robustness.

Raven’s Progressive Matrices (RPM) is highly correlated with human intelligence, and it has been widely used to measure the abstract reasoning ability of humans. In this paper, to study the abstract reasoning capability of deep neural networks, we propose the first unsupervised learning method for solving RPM problems. Since the ground truth labels are not allowed, we design a pseudo target based on the prior constraints of the RPM formulation to approximate the ground-truth label, which effectively converts the unsupervised learning strategy into a supervised one. However, the correct answer is wrongly labelled by the pseudo target, and thus the noisy contrast will lead to inaccurate model training. To alleviate this issue, we propose to improve the model performance with negative answers. Moreover, we develop a decentralization method to adapt the feature representation to different RPM problems. Extensive experiments on three datasets demonstrate that our method even outperforms some of the supervised approaches. Our code is available at https://github.com/visiontao/ncd .

In recent years, person re-identification (re-ID) has achieved relatively good performance, benefiting from the revival of deep neural networks. However, due to the existence of domain bias which refers to the different data distributions between two domains, it remains challenging to directly deploy a model trained on a labeled source domain to a target domain only with unlabeled data available. In this paper, a Self-Training with Progressive Representation Enhancement (PREST) framework, which comprises a multi-scale self-training method and a view-invariant representation learning module, is proposed to promote re-ID performance on the target domain in an unsupervised manner. More specifically, multi-scale representations, including the global body and local parts of pedestrian images, are utilized to obtain pseudo-labels. Then, some images are selected according to the pseudo-labels to create a new dataset for supervising the fine-tuning process, which is operated iteratively to progressively promote the performance. Furthermore, to mitigate the influence of different styles among sub-domains, in cases where a single sub-domain is captured by one camera, a classifier with a gradient reverse layer is first employed to learn view-invariant representation for pedestrian images with the same identity taken by different cameras; this can further enhance the reliability of the predicted labels and improve the cross-domain re-ID performance. Extensive experiments on three large-scale re-ID datasets demonstrate that our framework achieves significantly better performance than existing approaches.

Insertions and deletions (INDELs) comprise a significant proportion of human genetic variation, and recent papers have revealed that many human diseases may be attributable to INDELs. With the development of next-generation sequencing (NGS) technology, many statistical/computational tools have been developed for calling INDELs. However, there are differences among those tools, and comparisons among them have been limited. In order to better understand these inter-tool differences, five popular and publicly available INDEL calling tools-GATK HaplotypeCaller, Platypus, VarScan2, Scalpel, and GotCloud-were evaluated using simulation data, 1000 Genomes Project data, and family-based sequencing data. The accuracy of INDEL calling by each tool was mainly evaluated by concordance rates. Family-based sequencing data, which consisted of 49 individuals from eight Korean families, were used to calculate Mendelian error rates. Our comparison results show that GATK HaplotypeCaller usually performs the best and that joint calling with Platypus can lead to additional improvements in accuracy. The result of this study provides important information regarding future directions for the variant detection and the algorithms development.

Powered knee and ankle prostheses can perform a limited number of discrete ambulation tasks. This is largely due to their control architecture, which uses a finite-state machine to select among a set of task-specific controllers. A non-switching controller that supports a continuum of tasks is expected to better facilitate normative biomechanics. This paper introduces a predictive model that represents gait kinematics as a continuous function of gait cycle percentage, speed, and incline. The basis model consists of two parts: basis functions that produce kinematic trajectories over the gait cycle, and task functions that smoothly alter the weight of basis functions in response to task. Kinematic data from ten able-bodied subjects walking at twenty-seven combinations of speed and incline generate training and validation data for this data-driven model. Convex optimization accurately fits the model to experimental data. Automated model order reduction improves predictive abilities by capturing only the most important kinematic changes due to walking tasks. Constraints on range of motion and jerk ensure the safety and comfort of the user. This model produces a smooth continuum of trajectories over task, an impossibility for finite-state control algorithms. Random sub-sampling validation indicates basis modeling predicts untrained kinematics more accurately than linear interpolation.

With increased use of electronic medical records (EMRs), data mining on medical data has great potential to improve the quality of hospital treatment and increase the survival rate of patients. Early readmission prediction enables early intervention, which is essential to preventing serious or life-threatening events, and act as a substantial contributor to reduce healthcare costs. Existing works on predicting readmission often focus on certain vital signs and diseases by extracting statistical features. They also fail to consider skewness of class labels in medical data and different costs of misclassification errors. In this paper, we recur to the merits of convolutional neural networks (CNN) to automatically learn features from time series of vital sign, and categorical feature embedding to effectively encode feature vectors with heterogeneous clinical features, such as demographics, hospitalization history, vital signs, and laboratory tests. Then, both learnt features via CNN and statistical features via feature embedding are fed into a multilayer perceptron (MLP) for prediction. We use a cost-sensitive formulation to train MLP during prediction to tackle the imbalance and skewness challenge. We validate the proposed approach on two real medical datasets from Barnes-Jewish Hospital, and all data is taken from historical EMR databases and reflects the kinds of data that would realistically be available at the clinical prediction system in hospitals. We find that early prediction of readmission is possible and when compared with state-of-the-art existing methods used by hospitals, our methods perform significantly better. For example, using the general hospital wards data for 30-day readmission prediction, the area under the curve (AUC) for the proposed model was 0.70, significantly higher than all the baseline methods. Based on these results, a system is being deployed in hospital settings with the proposed forecasting algorithms to support treatment.

With the development of neuroimaging techniques, a growing amount of multi-modal brain imaging data are collected, facilitating comprehensive study of the brain. In this paper, we jointly analyzed functional magnetic resonance imaging (fMRI) collected under different paradigms in order to understand cognitive behaviors of an individual. To this end, we proposed a novel multi-view learning algorithm called structure-enforced collaborative regression (SCoRe) to extract co-expressed discriminative brain regions under the guidance of anatomical structure of the brain. An advantage of SCoRe over its predecessor collaborative regression (CoRe) lies in its incorporation of group structures in the brain imaging data, which makes the model biologically more meaningful. Results from real data analysis has confirmed that by incorporating prior knowledge of brain structure, SCoRe can deliver better prediction performance and is less sensitive to hyper-parameters than CoRe. After validation with simulation experiments, we applied SCoRe to fMRI data collected from the Philadelphia Neurodevelopmental Cohort and adopted the scores from the wide range achievement test (WRAT) to evaluate an individual’s cognitive skills. We located 14 relevant brain regions that can efficiently predict WRAT scores and these brain regions were further confirmed by other independent studies.

Due to the complexity of myocardial infarction (MI) waveform, most traditional automatic diagnosis models rarely detect it, while those able to detect MI often require high computing and storage capacity, rendering them unsuitable for portable devices. Therefore, in order for convenient real-time MI detection, it is essential to design lightweight models suitable for resource-limited portable devices. This paper proposes a novel multi-channel lightweight model (ML-Net), that provides a new solution for portable detection devices with limited resources. In ML-Net, each electrocardiogram (ECG) lead is assigned an independent channel, ensuring data independence and preserve the ECG characteristics of different angles represented by different leads. Moreover, convolution kernels of heterogeneous sizes are utilized to achieve accurate classification with only a small amount of lead data. Extensive experiments over actual ECG data from the PTB diagnostic database are conducted to evaluate ML-Net. The results show that ML-Net outperforms comparable schemes in diagnosing MI, and it requires lower computational cost and less memory, so that portable devices can be more widely used in the field of Internet of Medical Things(IoMT).

Many learning algorithms such as kernel machines, nearest neighbors, clustering, or anomaly detection, are based on the concept of 'distance' or 'similarity'. Before similarities are used for training an actual machine learning model, we would like to verify that they are bound to meaningful patterns in the data. In this paper, we propose to make similarities interpretable by augmenting them with an explanation in terms of input features. We develop BiLRP, a scalable and theoretically founded method to systematically decompose similarity scores on pairs of input features. Our method can be expressed as a composition of LRP explanations, which were shown in previous works to scale to highly nonlinear functions. Through an extensive set of experiments, we demonstrate that BiLRP robustly explains complex similarity models, e.g. built on VGG-16 deep neural network features. Additionally, we apply our method to an open problem in digital humanities: detailed assessment of similarity between historical documents such as astronomical tables. Here again, BiLRP provides insight and brings verifiability into a highly engineered and problem-specific similarity model., Comment: 12 pages, 10 figures

It has been a key topic to decompose the brain's spatial/temporal function networks from 4D functional magnetic resonance imaging (fMRI) data. With the advantages of robust and meaningful brain pattern extraction, deep neural networks have been shown to be more powerful and flexible in fMRI data modeling than other traditional methods. However, the challenge of designing neural network architecture for high-dimensional and complex fMRI data has also been realized recently. In this paper, we propose a new spatial/temporal differentiable neural architecture search algorithm (ST-DARTS) for optimal brain network decomposition. The core idea of ST-DARTS is to optimize the inner cell structure of the vanilla recurrent neural network (RNN) in order to effectively decompose spatial/temporal brain function networks from fMRI data. Based on the evaluations on all seven fMRI tasks in human connectome project (HCP) dataset, the ST-DARTS model is shown to perform promisingly, both spatially (i.e., it can recognize the most stimuli-correlated spatial brain network activation that is very similar to the benchmark) and temporally (i.e., its temporal activity is highly positively correlated with the task-design). To further improve the efficiency of ST-DARTS model, we introduce a flexible early-stopping mechanism, named as ST-DARTS+, which further improves experimental results significantly. To our best knowledge, the proposed ST-DARTS and ST-DARTS+ models are among the early efforts in optimally decomposing spatial/temporal brain function networks from fMRI data with neural architecture search strategy and they demonstrate great promise.

Deep learning based models have excelled in many computer vision tasks and appear to surpass humans' performance. However, these models require an avalanche of expensive human labeled training data and many iterations to train their large number of parameters. This severely limits their scalability to the real-world long-tail distributed categories, some of which are with a large number of instances, but with only a few manually annotated. Learning from such extremely limited labeled examples is known as Few-shot learning (FSL). Different to prior arts that leverage meta-learning or data augmentation strategies to alleviate this extremely data-scarce problem, this paper presents a statistical approach, dubbed Instance Credibility Inference (ICI) to exploit the support of unlabeled instances for few-shot visual recognition. Typically, we repurpose the self-taught learning paradigm to predict pseudo-labels of unlabeled instances with an initial classifier trained from the few shot and then select the most confident ones to augment the training set to re-train the classifier. This is achieved by constructing a (Generalized) Linear Model (LM/GLM) with incidental parameters to model the mapping from (un-)labeled features to their (pseudo-)labels, in which the sparsity of the incidental parameters indicates the credibility of the corresponding pseudo-labeled instance. We rank the credibility of pseudo-labeled instances along the regularization path of their corresponding incidental parameters, and the most trustworthy pseudo-labeled examples are preserved as the augmented labeled instances. Theoretically, under mild conditions of restricted eigenvalue, irrepresentability, and large error, our approach is guaranteed to collect all the correctly-predicted instances from the noisy pseudo-labeled set., Journal extension of arXiv:2003.11853 that appears in CVPR 2020 Code and models are released at https://github.com/Yikai-Wang/ICI-FSL