Lawrence H. Schwartz, MD

The goal of a project undertaken in 1998-1999 by the Department of Radiology at Memorial Sloan-Kettering (MSK) Cancer Center, New York City, was to create a digital archive of the medical images produced in the department. At the time, Memorial was opening a large outpatient center located 15 blocks away from its main hospital campus. There was a need for patients to be seen along with their imaging studies in both facilities. It was believed that a digital archive and distribution could improve the efficiency of this process such that the time for a radiology procedure may be shortened, the time of the images to be available would almost certainly decrease, and, potentially, the time that a radiology report would be available would be decreased as well. The department determined that the best means of achieving its aim to provide efficient service were digital data acquisition, digital data distribution, and the use of advanced data-analysis tools. Meeting these three objectives might reasonably be expected, in turn, to help the department achieve its goal of improving diagnostic accuracy.

A picture archiving and communications system (PACS) was identified as the technology most able to unify the department’s pursuit of its objectives and support the institution in reaching its goals. A preliminary step in pursuing PACS implementation consisted of reviewing the PACS needs and expectations of clinicians, radiologists, and the enterprise as a whole. From the clinician’s perspective, the PACS is largely invisible, being one of several integrated systems providing data at the workstation. For example, the clinician requires access to the clinical database, admission-discharge-transfer information, the hospital’s scheduling system, laboratory reports, and the billing system, in addition to radiology images and reports.

The general PACS features deployed by the institution were made available to oncologists in December, 1998; all imaging modalities feed data into the system. Fundamentally, the Digital Imaging and Communications in Medicine standard was employed so that all acquisition devices could communicate fully, and digital images and reports could be retrieved on demand by radiologists and clinicians. The PACS features most useful to oncologists and oncologic surgeons included the ability to compare a current study with a prior examination and to compare multiple imaging modalities on a single- or dual-monitor screen. It is now possible for an oncologist or surgeon to review a CT, MRI, and PET scan simultaneously and reference findings on one examination with another. In order to accomplish this goal, MSK decided to digitally migrate approximately 2 years’ worth of CT and MRI images on magneto-optical disks into the PACS.

Nonradiologist clinicians use PACS extensively to review current imaging studies during outpatient visits. Our analysis has demonstrated that busy clinicians will review images on PACS between 50% and 80% of outpatient visits. They will frequently compare these studies with older examinations to assess for interval change.


The need of oncologists for access to prior studies is clear. This need is more intense than in many other practices. For example, the recall rate in an average emergency department after 2 weeks is relatively low. The recall rate in a clinician’s office, such as an orthopedic surgeon, is moderate, but is very high in an oncology practice where it is hoped that patients return for follow-up visits and those with solid tumors generally have their disease assessed with new imaging studies. For this reason, digital image storage is a great advantage. Many PACS consist of a two tier archivea deep archive holds older or less accessed imagesand retrieval from the “deep” archive is generally on the order of minutes, in comparison to the arrays of hard drives with rapid access in seconds. Many centers, ourselves included, are in the process of building up the rapid retrieval to handle a greater capacity of studies, thus optimizing system performance.

In order to promote efficient use of the PACS user’s time, the system attempts to predict which studies will be needed from the archive. This permits them to be moved from the archive to disk storage so they can be viewed quickly. This process, called prefetching, typically uses a patient’s upcoming examination as the prompt to retrieve the relevant prior images according to protocols set in advance. For example, a clinic appointment may trigger a different level of prefetching than a radiology visit would; users may also determine how many studies will be prefetched and whether they will limit prefetching to certain modalities, body parts, or reasons for imaging (such as preoperative examinations). The length of time that prefetched studies will remain available may also be predetermined.

If a patient’s visit has been scheduled in advance, prior examinations can be prefetched by the PACS during the evening preceding the appointment. This protocol helps improve data-network traffic because fewer clinicians are likely to request images at this time. A more urgent visit that takes place on the same day that it was scheduled, however, will still trigger the prefetching of prior examinations. For oncology clinic appointments, we have found that prefetching the previous year’s examinations is satisfactory; for radiology appointments, the previous six examinations of the matching body part are prefetched. As rapid access, classically considered short-term storage is increased, some of these rules and paradigms will have to change.

The availability of prior imaging studies for comparison purposes is essential to the success of a clinical PACS. The historical archive of imaging studies may be developed naturally over time as digital images accumulate, or it may be created all at once through migration of existing digital images from their storage media to the PACS archive. The cost of allowing the archive to grow naturally may be significant, primarily because the previous storage and retrieval systems for images outside the PACS archive must be maintained. In addition, physicians are likely to be frustrated by the need to obtain prior images from two different systems. Confusion is also to be expected when prior studies are needed in order to perform new radiology studies. For these reasons, it is preferable to move as many stored images as possible to the PACS.

At Memorial Sloan-Kettering Cancer Center, migration to the PACS was performed for 49,985 CT and MRI examinations. The success of migration of this digital data was due in part to the ability to match the image data with data in the radiology information system. Matching this data was not trivial. In 40% of cases, matching of patient information was found to be inaccurate due to typographical errors or varying use of abbreviations in patient data. Nonetheless, data migration was believed to be beneficial and economical because it diminished film-library requests and accelerated the acceptance of PACS by radiologists and referring physicians.


Work lists automate the process of identifying patient studies for review. PACS work lists prevent reading of the same study by two radiologists; they also help distribute work appropriately (based on the subspecialties and work loads of the individuals who will interpret the images). Departments can set protocols that will be used by the system to meet their image-distribution needs and preferences. The output of multiple modalities or devices can be combined in one work list; likewise, multiple radiologists may chose to share one work list, thus facilitating image interpretation by residents and fellows, if necessary. Work lists have facilitated subspecialty care and expert consultation, even with radiologists spread across multiple campuses.

Complete connectivity is the invisible element that makes a constantly updated work list possible. The system must have access, in real time, to information concerning patient status, ordered imaging studies, upcoming clinic visits, clinical history (as needed for image interpretation), and verified radiology reports for clinical review.


As imaging studies are more rapidly available to clinicians for review, it is also essential for the radiologist’s report to be available as well. Automated speech recognition (ASR) systems recognize continuous speech with a high level of accuracy, and allow for more rapid distribution of the dictated report. Ideally, speech recognition is speaker independent and includes a vocabulary of 20,000 to 70,000 words. The potential benefits of ASR are reduced costs, decreased time between image acquisition and report distribution, and more rapid availability to clinics of images with reports (through ASR’s compatibility with PACS). The economic arguments favoring ASR are powerful. For example, a facility that generates 100,000 imaging reports per year and pays transcriptionists an average of $2 per report will spend $200,000. The same reports, if generated by 20 radiologists using ASR at 20 workstations, will initially cost $10,000 per workstation and $50,000 for training, or $250,000. This system may therefore pay for itself in a short period of time.

Of course, some prerequisites must be met before ASR can be adopted. Radiologists must be motivated to change their work habits, since they will be expected to assume the new tasks of proofreading and correcting their reports. While some radiologists will always turn in error-free reports, others will need to expend considerable effort to acquire this skill. For this reason, the facility must be willing to accept slower reporting during the initial phases of ASR implementation.


When a fully integrated PACS is in place and archival access, customized work lists, and ASR are combined with a disease-management approach, real work-flow automation becomes possible. PACS has complemented a disease management approach to oncology care at Memorial Sloan-Kettering Cancer Center. Subspecialists in oncology may review imaging studies in any location in the hospital enterprise. Radiologists also with subspecialty care may review imaging studies for their clinical colleagues again wherever they may be in the enterpriseeven if physically separated by distances. Patients can travel through the system without the need for their film jacket to go with them. Imaging studies may be displayed in operating rooms or at clinical conferences with ease and assurance that the images will be available. It is this type of efficiency that helps promote better, interactive, and informed patient care decisions.?

This article has been adapted from PACS in Oncology Practice, which he presented at the 88th annual meeting of the Radiological Society of North America on December 3, 2002, in Chicago.

Lawrence H. Schwartz, MD, is a radiologist in the Department of Radiology, Memorial Sloan-Kettering Cancer Center, New York City.