PARMED to provide user-friendly distributed video storage and retrieval facility for medical tele-education

Amsterdam 10 May 2000During the last Conference on High Performance Computing and Networking, the Polish-Austrian PARMED project was highlighted. The team is developing a Multimedia Storage and Retrieval System (MMSRS) for the remote delivery and sharing of digital images and video sequences between medical centres all over Europe, basically to address tele-educational and scientific purposes. Currently, the contents structure of the object-relational meta-database features visual data on three different surgical interventions, which are hip joint replacement, lumbar interbody fusion, and screwing in of the titanium block. The content of the image and video fragments is provided with annotation semantics, divided into the four classes of persons, events, objects, and locations. This approach has to facilitate the query and retrieval process for the user.


Three major challenges are to be met by the PARMED team when building the MMSRS database. First, the researchers have to provide enough capacity to store Terabytes of MPEG files. Second, the users should have no difficulty to efficiently extract video fragments from MPEG files, and thirdly, quality of service needs to be guaranteed. The huge amount of data requires tertiary storage in an automated tape library (ATL 4/52), equipped with the suitable UniTree Central File Manager (UCFM) software. The visual material is offered by the Cracow Rehabilitation Centre and includes a wide range of typical and non-typical medical cases. The team is working on two different approaches to design the MMSRS. The first implies the use of commercially available Hierarchical Storage Management (HSM) software as to control the mass storage hardware. The other consists in building a specialised storage management system from scratch.

The HSM approach requires the enhancement of the system functionality so that the database can easily store videos and access fragments of them. This is possible via the MPEG Extension for HSM (MEH). The video file queried by the user is first staged or copied to the Disk Cache (DC) where it can be read from. The consulted files are removed or purged from the DC through a Least Recently Used (LRU) algorithm. Since a user rarely needs to download a full video and in order to avoid start-up delay for really big files, the researchers have cut the videos into pieces of similar size to store them as separate files with the MPEG Store Application (MSA). Fragment retrieval is performed via the MPEG Retrieve Application (MRA). To this end, the user has to enter the name of the video as well as the frame range including start and end frames, into a "WWW" page using "HTTP" through a Common Gateway Interface (CGI).

The PARMED project team has decided to utilise a Proxy Cache Video Server (PCVS) to provide the user with the required quality of service (QoS). In many cases, this infrastructure is already available within university and hospital departments. The user-friendly PCVS approach allows to overcome problems of unreliable network bandwidth and to ensure smooth video delivery. PCVS has the additional advantage that the cached video fragments can be applied for further processing or interactive applications with repetitive access to the same set of videos. The fragments residing in the cache are replaced with the LRU method which will be expanded with a quality adaptation strategy in the future. The concepts and techniques developed in the PARMED project also can be adopted by non-medical collaborative entities.

Leslie Versweyveld

[Medical IT News][Calendar][Virtual Medical Worlds Community][News on Advanced IT]