Network monitoring and analysis in a large scale PACS and teleradiology environment

In addition to a departmental-wide PACS, UCLA is launching a nation-wide and international teleradiology project. In such a complex and geographically wide-spread networking environment, network management tools for system monitoring are critical for smooth operation. Earlier approaches either managed the system at the physical level, or monitored the network based on engineering objects rather than medical components. Another approach relied on PACS network modeling and simulation. At this stage, it is nearly impossible to model a PACS system completely and accurately, therefore development in network monitoring, analysis, tuning and upgrading of real systems becomes very important.; The main research direction is to develop an intelligent, centrally controlled real time monitoring system to evaluate PACS characteristics and performance parameters. Specifically, the purpose of this dissertation is to design and implement a monitoring system based on network management and object-oriented concepts. The system provides a friendly user graphical interface with the capability of visualizing high level medical entities such as scanners, patients, images and queues. This real time monitoring and analysis system was developed along three lines: technical aspects regarding system reliability and efficiency, operational aspects using the PACS war room, and clinical aspects involving the application of image retrieval estimation.; There were two types of monitoring: (1) a software approach which was SNMP monitoring based on the software platforms of IBM NetView 6000 and Harris Network Manager, and a customized agent for monitoring dynamic image flow; (2) a hardware approach which was achieved by a network sniffer device. A functional monitoring system was defined in terms of implementation, generation or extraction, processing, collection and presentation in an useful and intelligible format of dynamic information regarding the operation of a system.; A network management center, i.e. the PACS war room, with a graphical user interface allowed human experts to monitor and control different elements of the system through various levels of abstraction from one location. Examples in the PACS war room display include: hierarchical views of teleradiology and the PACS system, interactive structure of software processes, a dynamic image flow display, and network activity trends. Validation was performed in three ways: (1) spot checking, (2) problem handling and (3) informal user surveys. Short learning curve and efficient graphic visual effect was observed in the PACS war room.; A methodology was developed to predict the image retrieval delay based on a mathematical model. The results of estimation were compared with the true delays, and the model worked accurately within expected statistical variation except when the following situations occurred (1) transfers to designated research computers; (2) complex network topologies and long network paths to the destination computers, and (3) errors and retries due to multiple entries of the same image and catastrophic failure (e.g. disk full). Research machines should be isolated from the regular PACS operation to avoid serious system delay and to reduce the system complexity.