Irfan Habib

Providing traceability for neuroimaging analyses.

INTRODUCTION With the increasingly digital nature of biomedical data and as the complexity of analyses in medical research increases, the need for accurate information capture, traceability and accessibility has become crucial to medical researchers in the pursuance of their research goals. Grid- or Cloud-based technologies, often based on so-called Service Oriented Architectures (SOA), are increasingly being seen as viable solutions for managing distributed data and algorithms in the bio-medical domain. For neuroscientific analyses, especially those centred on complex image analysis, traceability of processes and datasets is essential but up to now this has not been captured in a manner that facilitates collaborative study. PURPOSE AND METHOD Few examples exist, of deployed medical systems based on Grids that provide the traceability of research data needed to facilitate complex analyses and none have been evaluated in practice. Over the past decade, we have been working with mammographers, paediatricians and neuroscientists in three generations of projects to provide the data management and provenance services now required for 21st century medical research. This paper outlines the finding of a requirements study and a resulting system architecture for the production of services to support neuroscientific studies of biomarkers for Alzheimers disease. RESULTS The paper proposes a software infrastructure and services that provide the foundation for such support. It introduces the use of the CRISTAL software to provide provenance management as one of a number of services delivered on a SOA, deployed to manage neuroimaging projects that have been studying biomarkers for Alzheimers disease. CONCLUSIONS In the neuGRID and N4U projects a Provenance Service has been delivered that captures and reconstructs the workflow information needed to facilitate researchers in conducting neuroimaging analyses. The software enables neuroscientists to track the evolution of workflows and datasets. It also tracks the outcomes of various analyses and provides provenance traceability throughout the lifecycle of their studies. As the Provenance Service has been designed to be generic it can be applied across the medical domain as a reusable tool for supporting medical researchers thus providing communities of researchers for the first time with the necessary tools to conduct widely distributed collaborative programmes of medical analysis.

Intelligent grid enabled services for neuroimaging analysis

This paper reports our work in the context of the neuGRID project in the development of intelligent services for a robust and efficient Neuroimaging analysis environment. neuGRID is an EC-funded project driven by the needs of the Alzheimers disease research community that aims to facilitate the collection and archiving of large amounts of imaging data coupled with a set of services and algorithms. By taking Alzheimers disease as an exemplar, the neuGRID project has developed a set of intelligent services and a Grid infrastructure to enable the European neuroscience community to carry out research required for the study of degenerative brain diseases. We have investigated the use of machine learning approaches, especially evolutionary multi-objective meta-heuristics for optimising scientific analysis on distributed infrastructures. The salient features of the services and the functionality of a planning and execution architecture based on an evolutionary multi-objective meta-heuristics to achieve analysis efficiency are presented. We also describe implementation details of the services that will form an intelligent analysis environment and present results on the optimisation that has been achieved as a result of this investigation.

Tools and Techniques for Managing Virtual Machine Images

Virtual machines can have many different deployment scenarios and therefore may require generation of multiple VM images. OS Farm is a service that aims to provide VM images that are tailored and generated on the fly. In order to optimize generation of images, a layered copy-on-write image structure is used, and an image cache ensures that identical images are not regenerated. Images can be several hundreds of megabytes large and thus can congest the network and delay their transfer. Content-Based Transfer is a technique which transfers only the difference between the source image and existing target client image data. We present an implementation which achieves an observed bandwidth close to the theoretical maximum and a significant reduction in network congestion.

From Grid Middleware to a Grid Operating System

Grid computing has made substantial advances during the last decade. Grid middleware such as globus has contributed greatly in making this possible. There are, however, significant barriers to the adoption of grid computing in other fields, most notably day-to-day user computing environments. We will demonstrate in this paper that this is primarily due to the limitations of the existing grid middleware which does not take into account the needs of everyday scientific and business users. In this paper we will formally advocate a grid operating system and propose an architecture to migrate grid computing into a grid operating system which we believe would help remove most of the technical barriers to the adoption of grid computing and make it relevant to the day-to-day user. We believe this proposed transition to a grid operating system will drive more pervasive grid computing research and application development and deployment in future

Provenance Management for Neuroimaging Workflows in neuGrid

An increased amount of large scale, collaborative biomedical research has recently been conducted on e-Science infrastructures. Such research typically involves conducting comparative analysis on large amounts of data to search for biomarkers for diseases. Running these analysis manually can often be quite cumbersome, labour-intensive and error-prone. Significant work has been invested into automating such analysis with appropriately configured workflows. It is also important for biomedical researchers to validate analysis outcomes, to ensure the reproducibility of the results and to ascertain the ownership of specific scientific results. The detailed, traceable information required for this is often referred to as provenance data. Developing suitable methods and approaches to managing provenance data in large-scale distributed e-Science environments is another important area of research currently being investigated. We present an approach that has been adopted in the neu GRID project, which aims to develop an infrastructure to facilitate research into neurodegenerative disease studies such as Alzheimers. To facilitate the automation of complex, large-scale analysis in neu GRID, we have adapted CRISTAL, a workflow and provenance tracking solution. The use of CRISTAL has provided a rich environment for neuroscientists to track and manage the evolution of both data and workflows in the neu GRID infrastructure.

A middleware agnostic infrastructure for neuro-imaging analysis

Large-scale neuroscience research projects are necessary in order to make significant progress in the study of degenerative brain diseases. At present the effectiveness of such efforts is being somewhat restricted by the absence of specifically tailored computing infrastructures. The neuGRID project aims to address this through the provision of a high-level service oriented infrastructure that enables complex neuro-science research. One of the principle aims of this work is to develop portable services that can be re-used in a larger set of related medical applications to access distributed computing resources. These services will provide high-level functionality that will support workflow authoring and planning, provenance storage and retrieval, querying against heterogeneous data sources as well as security and data anonymization amongst others. This paper introduces the neuGRID service architecture and outlines the design of two specific services, namely the Pipeline Service and the Glueing Service. A proof of concept implementation to evaluate the neuGRID design approach has been developed.

Adapting scientific workflow structures using multi-objective optimization strategies

Scientific workflows have become the primary mechanism for conducting analyses on distributed computing infrastructures such as grids and clouds. In recent years, the focus of optimization within scientific workflows has primarily been on computational tasks and workflow makespan. However, as workflow-based analysis becomes ever more data intensive, data optimization is becoming a prime concern. Moreover, scientific workflows can scale along several dimensions: (i) number of computational tasks, (ii) heterogeneity of computational resources, and the (iii) size and type (static versus streamed) of data involved. Adapting workflow structure in response to these scalability challenges remains an important research objective. Understanding how a workflow graph can be restructured in an automated manner (through task merge, for instance), to address constraints of a particular execution environment is explored in this work, using a multi-objective evolutionary approach. Our approach attempts to adapt the workflow structure to achieve both compute and data optimization. The question of when to terminate the evolutionary search in order to conserve computations is tackled with a novel termination criterion. The results presented in this article demonstrate the feasibility of the termination criterion and demonstrate that significant optimization can be achieved with a multi-objective approach.

Neuroimaging analysis using grid aware planning and optimisation techniques

Neuroimaging research is increasingly shifting towards distributed computing architectures for the processing of ever growing neuroimaging datasets. At present compute and data intensive neuroimaging workflows often use cluster-based resources to analyse datasets. For increased scalability however, distributed grid-based analysis platforms may be required. Such an analysis infrastructure necessitates robust methods of grid-aware planning and optimisation in order to efficiently execute often highly complex workflows. This paper presents the approaches used in neuGRID to plan the workflow gridification and enactment for neuroimaging research. Experiments show that grid-aware workflow planning techniques can achieve significant performance gains. Turn-around time of a typical neuroimaging workflow reduces by 30% compared to the same workflow enacted without grid-aware planning. Data efficiency also increases by more than 25%. The use of workflow planning techniques in the neuGRID infrastructure may enable it to process larger neuroimaging datasets and therefore allow researchers to carry out more statistically significant analysis.

A Service Oriented Analysis Environment for Neuroimaging Studies

neuGRID is an EC-funded project driven by the needs of the Alzheimers disease research community that aims to facilitate the collection and archiving of large amounts of imaging data coupled with a set of generalized services and algorithms. By taking Alzheimers disease as an exemplar, the neuGRID project has developed a set of generic analysis services and a Grid infrastructure that enables the European neuroscience community to carry out research required for the study of degenerative brain diseases. Using the services and the infrastructure, neuroscientists should be able to more easily identify neurodegenerative disease markers through the analysis of magnetic resonance and other brain imaging. The availability of such image-based disease markers will allow earlier diagnosis and foster the development of new drugs. This paper reports our work on the service oriented analysis environment that has been developed from a study of user requirements and that enables the neuroscience community to conduct and trace analyses for the study of Alzheimers and other neurodegenerative diseases. We present the salient features, architecture and implementation details of the services that will form an analysis environment. We also describe the functionality and benefits that these services will offer to the medical community in general and neuroimaging analysis in particular.

Rule-Based Querying of Distributed, Heterogeneous Data

Abstract When searching for data, users tend to think in terms of the information they need to retrieve and not where and how it is stored. This is especially true in the highly complex domain of neurological research. The generic medical querying service described herein, aims to close the gap between users and data resources that are intricate, distributed and heterogeneous in nature. A theme that is common to both this domain and the emerging Internet of things is that users often need to query more than a single resource. This has come about with the large-scale fragmentation and distribution of data. Our work thus far has produced a prototype architecture for the querying of heterogeneous distributed data in the medical domain. A flexible querying mechanism that has the potential to support agent-based personalization has been developed to provide users with accurate and relevant results and some initial results are presented.