Abstract

Introduction

Materials & Methods

Results

Discussion & Conclusion

References

Discussion
Board

The initial design and development of the PACS-browser took one man-month before being implemented. The elimination of films in CT and MRI at that time produced a significant dependence on the PACS-browser that brought to light various deficiencies. The most important of these was related to image quality which suffered from poorly selected window widths and levels being used to generate the JPEG versions of each image. This was rapidly addressed and also resolved similar issues with images being printed on paper. Several enhancements to the interface were also made over the first few months including the addition of more sophisticated database query tools and simple image manipulation controls.

As the user population grew and the volume of data managed by the PACS accumulated, we received feedback expressing concerns over image access times. In order to facilitate an investigation into these concerns, we conducted an analysis of the data volumes being handled on the network along with image transfer speeds². The results showed a high percentage of data older than three weeks being accessed primarily via the Web Server. This meant that users on the Web Server who needed these data were required to wait while the data were transferred (either from a Modality Server, or a Retrieval Server) and converted into JPEG format for viewing. This process also meant that there was a high degree of network traffic directed to the Web Server and that its CPU was computationally burdened with the overhead of DICOM communications and file format conversions.

At first, the solution seemed to be to increase the storage capacity of the Web Server by adding hard disks, but this would not resolve the issues of network and computational loads. Since the image data were distributed across multiple Servers (i.e., on the RAIDs of the Modality Servers), it became clear that it would be far more efficient to transfer the user to the data rather than transfer the data to the user. This was achieved simply by making each Modality Server a fully functional Web Server that could provide instant access to any images residing on its RAID.

The implementation of this plan required only minor modifications to the PACS-browser software and that each Modality Server generate a JPEG version of each image on its RAID (which is now done as each image is received from the scanner). The advantages of this modification include immediate access to any on-line image data on the PACS, and a more equitable distribution of computational load across the PACS. It also allowed the three weeks of storage provided by the four 9GB disks on the Web Server to be increased to over four months. In fact, once three of the 9GB disks were reallocated to the Modality Servers (since they were no longer required on the Web Server), this increased to nearly five months. A large reduction of network traffic was also achieved and the single point of failure that was previously inherent in the design was eliminated. Ultimately, this modification has allowed us to improve the service our PACS provides.

Although used primarily from within our institution, the Web-browser has also allowed us to extend the services of our department to users on the Internet. In anticipating this, we were careful to provide the means by which a user (eg, a referring physician) can easily obtain an account once authorized by their patient. Extensive logging is used throughout the system which facilitates audits of any user's actions and helps with any trouble-shooting.

A drawback to our scheme of distributing Web access across our Modality Servers is the difficulty of employing such an approach in conjunction with a firewall. Providing access to each Modality Server would require several points of access through the firewall which could increase vulnerability and be difficult to manage. Since a firewall will soon be installed at our institution, we will likely employ a single Web Server outside the firewall that would receive images from the PACS in a fashion similar to our original design, albeit only when necessary. Since there are fewer users outside the Hospital, we expect this approach will easily handle the loads.

Back to the top.