O

FIGURE 1 Configuration of an image archive system. The system consists of an archive server, a mirrored archive database, and a storage subsystem composed of cache storage (e.g., magnetic disks or high-speed RAID) and long-term storage (e.g. DLT tape library or optical disk library). The archive server is connected to a communication network, over which images are transmitted from the acquisition computers to the archive server and display workstations. A PACS gateway interfacing the Hospital Information System (HIS) and the Radiology Information System (RIS) allows the archive server to recieve information from HIS and RIS.

FIGURE 1 Configuration of an image archive system. The system consists of an archive server, a mirrored archive database, and a storage subsystem composed of cache storage (e.g., magnetic disks or high-speed RAID) and long-term storage (e.g. DLT tape library or optical disk library). The archive server is connected to a communication network, over which images are transmitted from the acquisition computers to the archive server and display workstations. A PACS gateway interfacing the Hospital Information System (HIS) and the Radiology Information System (RIS) allows the archive server to recieve information from HIS and RIS.

3.1 Archive Server

The archive server is a multitasking computer system that supports multiple processes (computer programs) to run simultaneously in its operation system environment. The archive server is configured with high-capacity RAM (random access memory), dual or multiple CPU (central processing unit), and high-speed network interface for better performance.

An integrated archive server runs sophisticated image management software that controls the archival, retrieval, and distribution of medical images for the archive system. Major functions performed by an archive server include (a) accepting images from the acquisition computers; (b) archiving images to the storage subsystem; (c) routing images to the display workstations; (d) updating archive database tables; and (e) handling query and retrieve requests from the display workstations.

3.2 Database Management Subsystem

The archive database is a relational database comprising predefined data tables that store information necessary for the archive server to perform the individual tasks supporting the image archive system.

To ensure data integrity, the archive database is configured to include a mirroring feature that allows data to be automatically duplicated on a separate system disk in the archive server. The archive database does not store any medical images. Instead, it stores the file index leading to access the corresponding images that are physically stored in the storage subsystem.

3.3 Storage Subsystem

The storage subsystem provides high-capacity storage for medical images and supports two levels of image storage: (1) short-term storage for data caching; and (2) long-term storage for permanent archiving.

Short-term storage uses fast data access storage devices such as magnetic disks or high-speed redundant array of inexpensive disks (RAID) that provide immediate access to images. Long-term storage, on the other hand, uses low-cost, high-capacity storage media such as magnetic tapes, optical disks, or digital linear tapes (DLT) that provide access to images at a slower speed.

3.4 Communication Network

The communication network is a digital interface that connects the archive server of the image archive system to other PACS components such as the acquisition computers and display workstations, allowing communications for medical images and relevant information.

The low-cost, 10-Mbps lOBaseT Ethernet can be used as the network interface to provide communications between the PACS components. However, high-bandwidth networks such as the 100-Mbps Fast Ethernet, 155-Mbps OC-3 ATM (asynchronous transfer mode), or 622-Mbps OC-12 ATM are more suitable for the communication because of the highvolume data transmission taking place in the PACS applications.

4 DICOM Image Communication Standard

Communication of images between medical imaging systems and among their applications has always been difficult because of the multiple platforms and vendor-specific communication protocols and data formats. The DICOM standard, developed in 1992 by a joint committee formed by the American College of Radiology (ACR) and the National Electrical Manufacturers Association (NEMA), is intended to provide connectivity and interoperability for multivendor imaging equipment, allowing communication of images and exchange of information among these individual systems [10]. This section describes two basic DICOM components (information objects and service classes) that are used for the communication of images.

4.1 Information Objects

Medical images are defined in DICOM as information objects or data sets. An information object represents an instance of a real-world information object (i.e., an image) and is composed of multiple data elements that contain the encoded values of attributes of that object. Each data element is made of three fields: the data element tag, the value length, and the value field. The data element tag is a unique identifier consisting of a group number and an element number in hexadecimal notation and is used to identify the specific attribute of the element. For example, the pixel data of an image is stored in the data element with a tag [7FE0, 0010], where 7FE0 represents the group number and 0010 represents the element number. The value length specifies the number of bytes that make up the value field of the element. The value field contains the value(s) of the data element. Figure 2 illustrates the composition of a DICOM image information object.

Image communication between medical imaging systems and among their applications takes place when a system or an application initiates a transfer of images to a designated system or application. The initiator (image sender) then transmits image data in terms of information objects to the designator (image receiver). In an image acquisition process, for example, medical images are transmitted as information objects from an image modality (e.g., a CT scanner) to a PACS acquisition computer. From the acquisition computer, these information objects are routed to their designated workstations for instantaneous display and to the archive server for archiving.

4.2 Service Classes

PACS applications are referred by DICOM as application entities (AEs). An AE that involves the communication of images is built on top of a set of DICOM services. These

FIGURE 2 DICOM information object. A DICOM image is an information object consisting of multiple data elements. Each data element is uniquely identified by its corresponding tag composed of a group number and an element number. Pixel data of the image is stored in element 0010 within group 7FE0.

services, performed by the DICOM message service elements (DIMSEs), are categorized into two service classes, the DIMSE-C services and the DIMSE-N services.

DIMSE-C services refer to those services that are applicable to composite information objects (i.e., objects that represent several entities in a DICOM information model) and provide only operation services. DIMSE-N services, on the other hand, are the services that are applicable to normalized information objects (i.e., objects that represent a single entity in a DICOM information model) and provide both operation and notification services. The DIMSE-C and DIMSE-N services and their operations are given in Tables 2 and 3, respectively.

A typical DIMSE service involves two AEs, a service class user (SCU), and a service class provider (SCP). A SCU is an AE that requests a specific DIMSE service from another AE (SCP). A SCP is an AE that performs an appropriate operation to provide a specific service. Operations carried out by the DIMSE services are based on client/service applications with the SCU being a client and the SCP being a server. Section 4.3 is an example of the storage service class that uses the C-STORE DIMSE service for transmitting medical images from a storage SCU to a storage SCP.

4.3 Example of C-STORE DIMSE Service

The following procedures describe the operation of a C-STORE DIMSE service that transmits images from a PACS acquisition computer to an archive server (Fig. 3):

(a) The acquisition computer (Storage SCU) issues an ASSOCIATION request to the archive server (Storage SCP)

(b) The archive server grants the association

(c) The acquisition computer invokes the C-STORE service and requests the storage of an image in the archive system

(d) The archive server accepts the request

(e) The acquisition computer transmits the image to the archive server

(f ) The archive server stores the image in its storage device and acknowledges successful operation

(g) The acquisition computer issues a request to drop the association

(h) The archive server drops the association

5 Archive Software Components

The software implemented in an archive server controls the archival, retrieval, and distribution of medical images for the archive system. In the archive server, processes of diverse functions run independently and communicate simultaneously

TABLE 2 Composite DICOM message service elements (DIMSE-C)

DIMSE-C service Operation

C-ECHO Verification of communication between two peer application entities (AEs)

C-STORE Transmission of information objects from one AE to another

C-FIND Querying information about the information objects

C-GET Retrieval of stored information objects from another AE using the C-STORE operation

C-MOVE Instructing another AE to transfer stored information objects to a third-party AE using the C-STORE operation with other processes by using client/server programming, queuing mechanisms, and job prioritizing mechanisms. Figure 4 illustrates the interprocess communication among the major processes running on the archive server, and Table 4 describes the functions of these individual processes.

The major tasks performed by the archive server include image receiving, image routing, image stacking, image archiving, database updating, image retrieving, and image pre-fetching. This section and Section 6.2 (Prefetch Mechanism) describe these individual tasks.

5.1 Image Receiving

Images acquired by the acquisition computers from various medical imaging devices are transmitted over the communication network to the archive server using standard DICOM storage service class via TCP/IP (transmission control protocol/ Internet protocol) network protocols. The storage service class is based on client/server applications, of which an acquisition computer (client) serves as a DICOM service class user (SCU) transmitting images to a service class provider (SCP) or the archive server. Like most client/server applications, the archive server supports concurrent connections to receive images from multiple acquisition computers.

5.2 Image Routing

Images arriving in the archive server from various acquisition computers are immediately routed to their destination display workstations for instantaneous display. The routing process is a DICOM storage service class, of which the archive server takes on the role of SCU to transmit images to the display workstations (SCP).

The routing algorithm is driven by predefined parameters such as examination type, patient type, location of display workstation, section radiologist, and referring physician. These parameters are stored in a routing table managed by the archive database. The routing process performs table lookup for each individual image based on the Health Level Seven (HL7) message received via the Hospital Information System/ Radiology Information System (HIS/RIS) interface (Section 6). Results from the table lookup will determine what destination(s) an image should be sent to.

5.3 Image Stacking

Image stacking is a data caching mechanism that stores images temporarily in high-speed storage devices such as magnetic disks and RAIDs for fast retrieval. Images received in the archive server from various acquisition computers are stacked in the archive server's cache storage (magnetic disks or RAID) to allow immediate access. After being successfully archived to the long-term storage, these images will remain in the cache storage and are managed by the archive server on a per-patient-per-hospital stay basis. In this way, all recent images that are not already in a display workstation's local disks can be retrieved from the archive server's cache storage instead of the low-speed long-term storage device such as an optical disk library or a DLT tape library. This timely access to images is particularly convenient for physicians or radiologists to retrieve images at a display workstation located in a different radiology section or department.

TABLE 3 Normalized DICOM message service elements (DIMSE-N)

DIMSE-N service

Operation

0 0

Post a comment