Search

Digital technologies, such as Big Data analytics, artificial intelligence, cloud and high-performance computing are presenting new opportunities to transform healthcare systems, increase connectivity of hospitals and other providers, and therefore potentially and significantly improve patient care. However, such networked computing infrastructures also raise significant cybersecurity risks, especially in the healthcare domain, where protecting sensitive personal information is of paramount importance. Project ASCLEPIOS aims at strengthening the trust of users in cloud-based healthcare services by utilizing trusted execution environment and several modern cryptographic approaches such as attribute based encryption, searchable encryption, functional encryption to build a cloud-based e-health framework that protects users’ privacy, prevents both internal and external attacks, verifies the integrity of medical devices before application, and runs privacy-preserving data analytics on encrypted data. The project investigates modern encryption techniques and their combination in order to provide increased security of e-health applications that are then presented towards end-users utilizing a cloud-based platform. Although some topics such as security and privacy are already investigated through block-chain related technologies, it has been decided that the selected approaches would be more suitable for these particular challenges. In order to prototype its security services, ASCLEPIOS develops and deploys three large-scale healthcare demonstrators, provided by three leading hospitals from Europe. These demonstrators are rooted in the practice-based problems and applications provided by the project’s healthcare partners. The Amsterdam University Centers, University of Amsterdam, plans to improve stroke hyper-acute care through secure information sharing on a cloud computing platform to improve patient management. Additionally, they are also building prediction models to enable earlier discharge of patients from hospitals with lower risk factors. Charité Berlin plans to improve inpatient and outpatient sleep medication by remotely controlling the quality of the collected data and transferring it on-line for further analysis. Finally, the Norwegian Centre for e-health Research, University Hospital of North Norway is developing a system for privacy-preserving monitoring and benchmarking of antibiotics prescription of general practitioners. The common characteristics of these three scenarios are the increased demand for high levels of security in data transfer, storage and privacy preserving analytics on cloud infrastructures. In order deploy, operate and further develop these applications to increase their security with the ASCLEPIOS framework, a cloud computing testbed is being setup. The testbed uses state-of-the-art technologies for cloud application deployment and run-time orchestration in order to ensure the optimized deployment and execution of the demonstrator applications. As the data sources do not require the local execution (albeit in one case data may remain on the data source) of processing, there is no need for fog or edge computing, but the testbed is based on private OpenStack cloud computing infrastructures and utilizes the MiCADO framework which is compatible with different containers such as Docker and Kubernetes. The project started only recently, and currently it is in the early stages of systems design and specification. This presentation will provide a short introduction to the ASCLEPIOS project and its demonstrators and will present early results of the currently ongoing requirements specification and platform design processes.

The meaning of the term 'Big Data' is still subject to debate, in spite of being widely used in biomedical publications. This confusion in definition leads to missed opportunities for peers to exchange knowledge and practices. A better understanding of Big Data may help researchers to identify themselves with the Big Data community. In this study, we investigate the most distinguishing features in 'Big Data'-labelled publications by comparing them against publication without this label (non-Big Data), using text mining and machine learning methods. Furthermore, the usage of the term Big Data was analysed over time. Our models could successfully make a distinction between publications labelled with 'Big Data' and those without. The most distinguishing features consisted of terms such as 'omics', 'computing', 'storage', and 'mining'. We observed that publications that do not use the term Big Data may also address topics that fall under well-accepted definitions of Big Data. Trends suggest that, while the use of the term Big Data increased, it is used less reliably now as compared to earlier years.

Finding all relevant publications to perform a systematic review can be a time consuming task, especially in the field of diagnostic test accuracy. Therefore, the CLEF eHealth lab 'technologically assisted reviews in empirical medicine' was established to create a basis of comparison between various methods. In this paper we describe a method submitted to the lab. This method consists of a topic model used to extract features and a random forest to classify the relevant papers. Classifier performance shows and average decrease of 33.3% in workload (i.e., documents to read) when aiming for a 95% recall and 24.9% for 100% recall. However, there is a large variety in workload reduction (79.3% to 0.9%) between the diagnostic test accuracy reviews.

To support scientists of different disciplines, different fields of Computer Science have developed tools and infrastructures with the aim of giving them access to vast computational resources in the easiest possible way. Such extremely complex structures have evolved naturally in the last decades both in depth and breath and, in addition to scientists, a plethora of heterogeneous actors (system administrators, developers, etc.) cooperate and interact with them. This complex and unstructured flow of actions and information poses difficulties in the development and usage of Science Gateways because information can be missing or hard to isolate at the right layer. In this paper, we aim to start a discussion on how to best manage these information flows to help the design and implementation of more flexible and userfriendly Science Gateways and workflow management systems in the future.

Workflow management has been widely adopted by scientific communities as a valuable tool to carry out complex experiments. It allows for the possibility to perform computations for data analysis and simulations, whereas hiding details of the complex infrastructures underneath. There are many workflow management systems that offer a large variety of generic services to coordinate the execution of workflows. Nowadays, there is a trend to extend the functionality of workflow management systems to cover all possible requirements that may arise from a user community. However, there are multiple scenarios for usage of workflow systems, involving various actors that require different services to be supported by these systems. In this paper we reflect about the usage scenarios of scientific workflow management based on the practical experience of heavy users of workflow technology from communities in three scientific domains: Astrophysics, Heliophysics and Biomedicine. We discuss the requirements regarding services and information to be provided by the workflow management system for each usage profile, and illustrate how these requirements are fulfilled by the tools these communities currently adopt. This paper contributes to the understanding of properties of future workflow management systems that are important to increase their adoption in a large variety of usage scenarios.

International audience

International audience; The integration of grid, cloud and other e-infrastructures into the fields of biology, bioinformatics, biomedicine, and healthcare are crucial if optimum use is to be made of the latest high-performance and distributed computer technology in these areas. Science gateways are concerned with offering intuitive graphical user interfaces to applications, data, and tools on distributed computing infrastructures. This book presents the joint proceedings of the Tenth HealthGrid Conference and the Fourth International Workshop on Science Gateways for Life Sciences (IWSG-Life), held in Amsterdam, the Netherlands in May 2012. The HealthGrid conference promotes the exchange and debate of ideas, technologies and solutions likely to promote the integration of grids into biomedical research and health in the broadest sense. The IWSG-Life workshop series is a forum that brings together scientists from the field of life sciences, bioinformatics, and computer science to advance computational biology and chemistry in the context of science gateways. These events have been jointly organized to maximize the benefit from synergies and stimulate the forging of further links in joint research areas. HealthGrid Applications and Technologies Meet Science Gateways for Life Sciences is divided into three parts. Part I includes contributions accepted to the HealthGrid conference; Part II contains the papers about various aspects of the development and usage of science gateways for life sciences. The joint session is recorded in Part III, and addresses the topic of science gateways for biomedical research. This book will provide insights and new perspectives for all those involved in the research and use of infrastructures and technology for healthcare and life sciences.

Computational neuroscience is a new field of research in which neurodegenerative diseases are studied with the aid of new imaging techniques and computation facilities. Researchers with different expertise collaborate in these studies. A study requires scalable computational and storage capacity and information management facilities to succeed. Many virtual laboratories are proposed and developed to facilitate these studies, however most of them cover only the parts related to the computational data processing. In this paper we describe and analyse the phases of the computational neuroscience studies including the actors, the tasks they perform, and the characteristics of each phase. Based on these we identify the required properties and functionalities of a virtual laboratory that supports the actors and their tasks throughout the complete study.

This paper presents the integration of a workflow management system into the VL-e medical software architecture. Workflows are designed with the Taverna workbench, and then executed with the MOTEUR engine on the EGEE grid through the VBrowser, which is the basic front-end for grid-enabled applications in the VL-e medical project. Data management is handled by the virtual file system of the VL-e toolkit. The resulting system provides a high-level interface to execute grid applications.

Measuring the diffusion properties of crossing fibers is very challenging due to the high number of model parameters involved and the intrinsically low SNR of Diffusion Weighted MR Images. Noise filtering aims at suppressing the noise while pertaining the data distribution. We propose an adaptive version of the Linear Minimum Mean Square Error (LMMSE) estimator to achieve this. Our filter applies an adaptive filtering kernel that is based on a space-variant estimate of the noise level and a weight consisting of the product of a Gaussian kernel and the diffusion similarity with respect to the central voxel. The experiments show that the data distribution after filtering is still Rician and that the diffusivity values are estimated with a higher precision while pertaining an equal accuracy. We demonstrate on brain data that our adaptive approach performs better than the initial LMMSE estimator

Diffusion Tensor MRI (DTI) is a rather recent image acquisition modality that can help identify disease processes in nerve bundles in the brain. Due to the large and complex nature of such data, its analysis requires new and sophisticated pipelines that are more efficiently executed within a grid environment. We present our progress over the past four years in the development and porting of the DTI analysis pipeline to grids. Starting with simple jobs submitted from the command-line, we moved towards a workflow-based implementation and finally into a web service that can be accessed via web browsers by end-users. The analysis algorithms evolved from basic to state-of-the-art, currently enabling the automatic calculation of a population-specific ‘atlas’ where even complex brain regions are described in an anatomically correct way. Performance statistics show a clear improvement over the years, representing a mutual benefit from both a technology push and application pull.

Most existing Grid technology has been foremost designed with performance and scalability in mind. When using Grid infrastructure for medical applications, privacy and security considerations become paramount. This leads to a re-thinking of implementation and deployment aspects of common components of the current Grid architecture. This paper describes the impact of privacy and security considerations on the Grid infrastructure design, and enumerates trust aspects which must underpin the design of Grid technology to support medical applications. We propose a novel security framework for securely handling privacy sensitive information on the Grid.

Diffusion tensor imaging (DTI) provides insight into the white matter of the human brain, which is affected by schizophrenia. By comparing a patient group to a control group, the DTI-images are on average expected to be different for white matter regions. Principal component analysis (PCA) and linear discriminant analysis (LDA) are used to classify the groups. In this work, the number of principal components is optimised for obtaining the minimal classification error. A robust estimate of this error is computed in a cross-validation framework, using different compositions of the data into a training and a testing set Previously, sequential runs were performed in MATLAB, resulting in long execution times. In this paper we describe an experiment where this application was run on a grid with minimal modifications and user effort. We have adopted a service-based approach that autonomously launches image analysis services onto a campus-wide Condor pool comprising of volunteer resources. This allows high throughput analysis of our data in a dynamic resource pool. The challenge in adopting such an approach comes from the nature of the resources, which change randomly with time and thus require fault tolerance. Through this approach we have reduced the computation time of each dataset from 90 minutes to less than 10. A minimal classification error of 22% was obtained, using 15 principal components.

Grid workflows can produce thousands of results that should be properly organised to enable further analysis. Typically results are stored on locations hard-coded in the workflow or in the components, limiting reusability. In this paper we present an approach to (re)organise the output files generated by a grid workflow in a distributed storage environment. We propose to perform a post-mortem mapping of workflow results into a directory structure. This mapping is based on data provenance information and exploits grid catalog features, namely logical file names, to avoid data replication. By defining different mappings, users can generate their own semantic view of results generated during a workflow execution, which fosters user-friendliness, whereas preserving workflow reusability. An implementation on the Virtual Resource Browser (VBrowser) framework is detailed and evaluated on neuroimaging workflows. Results show that the complex directory structure of an image analysis application cane properly generated by our system. An initial performance evaluation of the mapping resolution and directory structure creation indicates that this approach provides a practical, simple, yet powerful solution to an important roadblock for the adoption of workflows to implement complex image analysis pipelines.

It is widely recognized that automatic segmentation is hard, leading to the state where user intervention cannot be avoided. In this paper we review existing literature and propose a systematic approach for the integration of automatic and interactive segmentation methods into one unified process. A framework and requirements for intelligent interactive segmentation are formulated, and an example is presented

Neuroimaging is a field that benefits from distributed computing infrastructures (DCIs) to perform data- and compute-intensive processing and analysis. Using grid workflow systems not only automates the processing pipelines, but also enables domain researchers to implement their expertise on how to best process neuroimaging data. To share this expertise and to promote collaborative research in neurosciences, ways to facilitate the exchange, re-use, and interoperability of workflow applications between different groups are required. The SHIWA project (SHaring Interoperable Workflows for large-scale scientific simulations on Available DCIs) is specifically addressing such use-cases, building a generic platform to facilitate workflow exchange and execution environments interoperability. The goal is to facilitate the dissemination and execution of workflows by diverse workflow management systems on multiple DCIs. This platform enables researchers to gain access to a variety of ready-to-use workflows, to reuse workflows developed by collaborators, to publish their own workflows to be used by others, and to use additional resources from external DCIs to run workflows. In this demonstration we present how the SHIWA platform is used to implement various usage scenarios in which workflow exchange supports collaboration in neuroscience. The SHIWA platform and the implemented solutions are presented from the user perspective, in this case the workflow developers and the neuroscientists. These workflow interoperability solutions aim to facilitate and enable more advanced and large-scale research in neuroscience. The demonstration will focus on usage scenarios currently employed to exchange, combine and interoperate neuroimaging workflows between Academic Medical Center, Amsterdam, the Charite Universittsmedizin, Berlin and the outGrid infrastructure. These workflows are developed for the analysis of neuroimaging data, in particular Magnetic Resonance Images (MRI) and Diffusion Tensor Imaging (DTI). Each group has ported workflows with complementary and overlapping functions to its Grid infrastructure using different workflow systems, so the goal is to combine and share them across the boundaries of the original DCIs. The following usage scenarios will be addressed in the demonstration: • Preparing the workflow for sharing with others (VO, workflow management system dependencies); • Using the SHIWA repository for publishing workflows to share a new workflow with other potential users; • Finding a workflow in the SHIWA repository and testing it with sample data; • Running the workflow found in the repository with own data using the SHIWA simulation platform; • Combining complementary workflows into a meta-workflow to implement additional functionality or combining different implementations of the same workflow to compute on different DCIs simultaneously The shown solution includes neuroimaging workflows using the GWES, MOTEUR, LONI pipeline and P-GRADE workflow engines submitting jobs to the German MediGRID, the Dutch BiG Grid, the European EGI, and the international outGrid infrastructures. Data to be accessed might be stored locally, on an iRODS data management system, in the LFC file catalog and on gridFTP-enabled sites. The user-interfaces are web-based and include a Liferay-based Grid portal and a P-GRADE workflow editor implemented as webstart application. The neuroimaging applications include self-developed tools for preprocessing DTI data and widely used methods from the ITK and the FSL toolboxes

This thesis presents various secure mechanisms for cross-organisational data sharing during acute stroke care. In such situations, access to the data is urgent. For all the proposals, we assume the patients' EMR are stored in a cloud system to improve accessibility and collection of medical records during the emergency. However, to protect against abuse or internal attacks in the cloud providers, the patient's EMR is stored as ciphertext in the cloud, and the encryption keys are only shared with the involved acute care teams. Additionally, the proposed protocols are used to dynamically grant and revoke access to the patient's EMR for the healthcare teams according to the needs of triage, diagnosis, hospital selection and treatment. This thesis presents mechanisms for secure cross-organisational data sharing during acute stroke care. The proposed mechanisms are applicable and generalisable to other acute and non-acute care cases. These mechanisms focus on improving data availability once the security requirements are fulfilled, so the professionals no longer need break-glass procedures, even in emergencies. Instead, access to data happens with lawful purpose, without compromising patient privacy. Furthermore, cross-organisational data sharing requires trustworthiness among organisations, which needs to be built with regulation compliance and transparency. The proposed mechanisms of this thesis may help future healthcare digital systems comply with the regulations and be more transparent, thus improving security and enabling data sharing across organisations.

Big Data is a term that has been around for many years and it is often understood as the use and manipulation of large volumes of data. Over the years more aspects of Big Data have been recognised by researchers and institutions, such as its velocity, variety, and value. Nowadays, many definitions exist and researchers have wildly different understandings of Big Data. We believe that, without an unambiguous definition, communication is hampered, resulting in missed opportunities for both the developers as well as users of Big Data technologies. A group of researchers that may benefit greatly from applying Big Data technologies to partially automate their work are systematic reviewers. The nature of their work involves reading, dissecting, and selecting considerable numbers of scientific papers. Over the years many tools have been developed to support reviewers. However, it is unclear how often these are used and how well they work. In this thesis we start with an exploration of a common understanding of the term Big Data. The focus then shifts to systematic reviewers and their use of tools to deal with Big Data. Lasty, we propose a new method that may improve these tools. In this thesis we aim to: uncover a common understanding of Big Data in the (bio)medical research field; aid in improving the adoption of automation tools among systematic reviewers; and, contribute to the effectiveness of automation tools.

This thesis aims to investigate the use of high-performance computing (HPC) techniques in developing imaging biomarkers to support the clinical workflow of acute stroke patients. In the first part of this thesis, we evaluate different HPC technologies and how such technologies can be leveraged by different image analysis applications used in the context of acute stroke care. More specifically, we evaluated how computers with multiple computing devices can be used to accelerate medical imaging applications in Chapter 2. Chapter 3 proposes a novel data compression technique that efficiently processes CT perfusion (CTP) images in GPUs. Unfortunately, the size of CTP datasets makes data transfers to computing devices time-consuming and, therefore, unsuitable in acute situations. Chapter 4 further evaluates the algorithm's usefulness proposed in Chapter 3 with two different applications: a double threshold segmentation and a time-intensity profile similarity (TIPS) bilateral filter to reduce noise in CTP scans. Finally, Chapter 5 presents a cloud platform for deploying high-performance medical applications for acute stroke patients. In Part 2 of this thesis, Chapter 6 presents a convolutional neural network (CNN) for detecting and volumetric segmentation of subarachnoid hemorrhages (SAH) in non-contrast CT scans. Chapter 7 proposed another method based on CNNs to quantify the final infarct volumes in follow-up non-contrast CT scans from ischemic stroke patients.

Cardiovascular-related diseases are a major cause of death worldwide. Stroke and cardiac arrest, are among the most common cardiovascular diseases. Prediction models for predicting the risk of complications and outcome of patients who suffer from a cardiovascular disease have been broadly explored. Given the heterogeneous and complex nature of medical data, most common prognostic models focus on a single type of data. This thesis focuses on the application of machine/deep learning for the field of cardiovascular-related diseases, while taking into account the information available in different data types. We developed and evaluated the accuracy or prediction models using a single type of data and the added value of including information from other data types. We also explored multiple combination strategies, using 2D and 3D image representations, Transfer Learning, autoencoders, ResNets, among others . Finally, we strived to make our models transparent by exploring multiple model visualization tools. We developed models for predicting the occurrence of delayed cerebral ischemia in patients who suffered from hemorrhagic stroke, for predicting functional outcome in patients who suffered from ischemic stroke, for predicting outcome after cardiac arrest, and for identifying patients with the Phospholamban gene mutation. This thesis shows that machine/deep learning can be successfully applied to multiple prediction and classification tasks in the field of cardiovascular diseases and can lead to significant improvements in prognosis accuracy. Moreover, it has been shown that combining multiple types of data can have a significant impact in model performance and lead to the discovery of new insights.

Biomedical researchers are facing data deluge challenges such as dealing with large volume of complex heterogeneous data and complex and computationally demanding data processing methods. Such scale and complexity of biomedical research requires multi-disciplinary collaboration between scientists from different organizations. Data-driven or e-Science methods are defined as a combination of Information Technology (IT) and science that enables scientists to tackle the data deluge challenges. The IT infrastructures that address these challenges are known as cyberinfrastructures or e-Infrastructures, which are the environments that provide collaborative sharing of distributed computing and data resources. However, e-Infrastructures fall short of high-level and customized services to support the needs of scientists genuinely, and scientists find interacting with e-Infrastructures challenging, as it requires detailed technical knowledge. Science Gateway (SG) research addresses these drawbacks. SGs are web-based enterprise information systems that provide scientists with customized and easy access to community-specific data collections, computational tools, and collaborative services on e-Infrastructures. In this thesis we advanced the understanding of the fundamentals of SGs for biomedical research by organizing the findings about the requirements of biomedical researchers. We also organized the considerations about the design, development, operation, and sustainability of effective SGs. Moreover, we constructed a few successful SGs that were adopted by a large number of scientists and facilitated their biomedical big data analysis on e-Infrastructures. Finally we proposed a reference model that organizes the essential functions of SGs. We think that these efforts will facilitate design, development, operation, sustainability, and most importantly, adoption of SGs for biomedical research.