Data Management & Workflow Improvements In Biomedical Imaging
Important aspects for an efficient and secure bioimaging workflow is a structured and secure data management system with focus on usability. Also integrated access to high performance computation and visualization is of interest in certain application areas. Important data such as images and signals acquired by different researchers are sometimes by default stored in shared local storage areas using the same user account. The mechanism commonly used for data transfer either between internal systems or between sites is most often a portable USB type storage device. No backup or data protection plans exists. This all together creates “data chaos” with an imminent risk for data loss and security related issues.
The aim of this project is to architect and implement a solid and secure data management infrastructure for biomedical image workflow with focus on usability and scalability. Biomedical sensors such as MRI, PET and SPECT using different hardware platforms and operating systems will be seamlessly integrated into a secure multiuser environment. One of the main design goals is to create a flexible and user-friendly environment enabling the users to perform their work in an intuitive way focusing on the scientific research instead of computer related issues. The system will include multiple levels of data protection and techniques for cluster and grid based computation and remote visualization will be evaluated.
The heterogeneous compute environment consists of a variety of hardware and software including proprietary systems and interfaces used for data acquisition. Five different operating systems are currently in use and all data stored must be accessible in a transparent way independent of operating system used.
The primary storage solution consists of a virtualized server environment running on three physical 2U servers. The main storage is based on three high availability RAID-6 disk arrays that provides fault tolerance from two drives failure i.e. two drives are allowed to fail without affecting the array. The total primary storage space is 66 TB. Nightly backups are performed to off-site servers with direct attached RAID-6 storage hosted in a separate building 1 km away from the imaging facility. User authentication and authorization is performed using a directory service and all data is stored privately for all users and research groups.
Techniques for parallel cluster based computation and remote visualization will be tested and evaluated. Typical areas of interest are reconstruction of tomography data and volumetric visualization.
PI: Prof. Freddy Ståhlberg
Dept. of Medical Radiation Physics, Lund University
Prof. Deniz Kirik
Brain Repair and Imaging in Neural Systems (BRAINS), Lund University
Prof. Göran Sandberg
Lunarc, Lund University
Tech Dir. Anders Follin
Lunarc, Lund University
Links and references
Lund Bioimaging Center: www.med.lu.se/bioimaging_center
Lunarc Center for Scientific and Technical Computing: www.lunarc.lu.se