Abstract
Recently, delivery service of DICOM (digital imaging and communications in medicine) images from a picture archiving and communication system (PACS) through intrahospital network has been considered to be necessary for the improvement of medical care in hospitals. To avoid effects of large burst data from PACS on other traffics such as data from hospital information systems (HIS) and radiology information systems (RIS), the bandwidth of the network for PACS service has been considered to be crucial. Thus, hospitals usually introduced high-grade network equipments to support PACS service. However, the necessity of broad banded intrahospital network is a bottleneck of the spreading of PACS among hospitals because these equipments are expensive and manpower consuming for maintenance.
In the present study, we analyzed network traffic for several months using SNMP and MRTG (Multi Router Traffic Grapher) in a hospital (having 1,000 beds) in which experimental PACS service had been supplied in addition to RIS and HIS (order entry service). The results of the experimental service suggest that the PACS traffic flow fits with the small world model of D. J. Watts. Thus, a suitable intrahospital network using OSPF (open shortest path first) routing protocol was constructed based on the small world model with fixed-port type small layer 3 switches in place of high grade network equipments. Under these conditions, we set 180 PACS clients in the hospital and started the PACS service as a regular service with a Kodak DIRECTVIEW PACS system. At first, we have evaluated time required from the initiation of downloading a DICOM image to the period at which image viewing can be started. Compressed and uncompressed DICOM images of CT and MRI were used for the evaluation. Capacities of the small layer 3 switches were high enough to handle the PACS traffic as revealed by traffic monitoring. The time required for uncompressed DICOM images was significantly shorter than that for the compressed images. Although this may be due to abilities of PACS servers and those of personal computers (PC) used as PACS clients in particular CPU clock and RAM size, the findings indicate that the bandwidth of the network is not a rate-limiting factor for the PACS service under the above-mentioned network conditions. This is in agreement with the data from traffic analysis, and thus, uncompressed DICOM images have been delivered successfully in the hospital.
On the basis of these findings, we have constructed an intrahospital network for a new hospital (with 600 beds) in which graded-up PACS service is available under film-less conditions with an electronic medical record system (having 730 clients). From the beginning of hospital service, nearly all PACS service using uncompressed DICOM images was started. At 1 month after hospital service started, this delivery service is practical use on clinical service.
