Disruptive technology takes radiologists out of their protective armor and forces them to become quick and nimble
The Battle of Crecy, fought in 1346, was the first major clash of the Hundred Years' War. The outnumbered English army led by Edward III defeated a vastly superior force of French and mercenary soldiers. The estimated 12,000 French casualties, compared with several hundred on the English side, outnumbered the entire English expedition. Similar results occurred in the better known 1415 Battle of Agincourt dramatized in Shakespeare's Henry V. The decisive English advantage in both battles was the longbow, a disruptive innovation developed by the Welsh in the 12th century.
This disruptive technology had three requirements for success: concept, design, and user expertise. Initial concept of the longbow is unknown, but by the early 14th century skilled Welsh craftsmen known as bowyers were creating weapons whose design leap-frogged existing weaponry. Extensive practice was necessary to achieve the skill needed to use the longbow effectively, and archers typically began training at age seven.
PACS is a disruptive technology with the same three requirements. The concept of soft-copy reading with rapid retrieval and display of images has been with us for over a decade. And just as the longbow design was critical to its success, PACS software must continue to evolve to allow increasing radiologist speed and accuracy. User training has overcome initial resistance but must respond to ongoing challenges.
Parenthetically, the Battle of Crecy is viewed by some as the end of chivalry. One unintended result of the long-distance effectiveness of the longbow was a decrease in close-quarter combat and the civil conventions then observed in warfare. Radiologists bemoaning diminished comity with their clinical colleagues likewise cite PACS and the physical disconnect created by the ability to view soft-copy images outside of the radiology department.
PACS as disruptive technology involves several issues under the concept, design, and user expertise umbrella.
- Work list functionality. Early PACS, as well as a number of current software versions, allowed basically two states on the radiologists' work list: read or unread. This can be quite restrictive in typical department workflow. Imagine, for example, reading from the electronic work list and coming across a case that needs comparison with archived film. You would leave this case on the work list, the digital equivalent of placing an unread film jacket back in the pile to be stumbled over by your colleagues. Newer work list functionality must allow the easy assignment of a case to a secondary work list. This can be used for communication with film library, technologists for quality improvement purposes, and radiologist colleagues when seeking a second opinion.
Most conventional PACS work lists can be sorted by time of study, stat reading, and other parameters. Dr. Kevin McEnery at M.D. Anderson Cancer Center has developed an innovative way to link the needs of patients and clinicians to the radiologist work list. His software enables the user to sort the work list based on patient clinic appointment time. This allows radiologists to ensure that report availability is timed to patient and clinician needs. This enterprise-centric approach greatly facilitates systemwide efficiency and decreases phone calls requesting a facilitated reading.
- Hanging protocols. Hanging protocols have continually increased in sophistication and usability with each software upgrade. Reading protocols are becoming available that will present a sequence of user-defined image display configurations. The capacity for automated distinction between PA and lateral chest radiographs should be included in new hanging protocols along with the capability of automatically segmenting body parts on a CT or MR scan.
PACS displays should also incorporate the data sets created by either native or third-party reconstruction or specialty workstations. Indeed, a number of radiologists are beginning to use the reconstruction workstations themselves as their primary image viewer. Any reporting software must be integrated with the images and must allow full user customization. Some radiologists find it useful, for example, to treat the dictation screen as one series of a CT or MR scan, keeping it adjacent to the images to prevent excessive eye movements during image interpretation and reporting.
- Graphical user interface. The GUI in some PACS software requires significant visual input, whether by keyboard search, tool palette, or pull-down menu. This must end. The eyes of the radiologist must be kept on the target at all times to allow maximum concentration and diagnostic accuracy. Imagine in the analog world reaching for the wax pencil, annotating a film, and then finding the pencil stuck to your hand. In soft-copy viewing, the choice of a particular single-use tool often becomes an unwanted default.
Certain tools that are frequently used in sequence should be easily linkable. For example, seamless toggle between pan and zoom functions must be facilitated in any efficient PACS software. The digital tool palette, in whatever form, must be completely configurable by the user and should be specific for each modality. The user must have the option of using any tool either one time only or as the default. A number of manufacturers are developing novel approaches to tool choice and manipulation with minimal distraction from image viewing.
- User interface devices. Most PACS use keyboard and mouse as the user interface. The typewriter and keyboard were invented in the 1870s by C. Latham Sholes, an unassuming printer from rural Pennsylvania. The provenance of the QWERTY keyboard configuration is often attributed to the requirement for smooth physical operation of the mechanical typewriter components, and it may, in fact, have been designed to slow down the typist to achieve this end. Its anachronistic use in PACS, comparable to the presence of a horse and buggy on the interstate, would be amusing were it not so detrimental to workflow.
The mouse was invented in the 1960s by Douglas Englebart, a true computer pioneer who also gave us the concepts of the modern GUI, hyperlink, and online conferencing. The problem with these devices is that they require three hands to be operated without visual input, an evolutionary feat that likely will not be achieved in our lifetimes. PACS software was designed around this combination simply because it existed. Newer mechanical user interfaces such as the multifunction mouse, gaming devices, and programmable combinations of jogwheel and button controllers are being used by some radiologists, but these also fall short of the mark.
- Integration. Whatever the user configuration and interface devices, the overall goal of the designer must be to maximize radiologist efficiency and allow access to all information from a single point of contact. This will require full integration of PACS with the RIS, reporting software, and electronic medical record. As the need for such software linkage is realized, previously well-defined delineations between these systems are blurring, and different combinations of software are being offered by vendors.
Integrating the Healthcare Enterprise profiles are already addressing many of the requirements for system integration, but the availability of so much data may result in further information overload as radiologists gain access to the entire patient record. To prevent this, the radiologist should receive just the amount of information necessary for image interpretation and decision making. This will ultimately involve sophisticated filtering of clinical data. Perhaps the simplest example of such filtering is a display of only the latest BUN and creatinine (rather than the entire lab report) for the radiologist when a decision is needed on whether to administer intravenous contrast for a CT scan. More complex algorithms will require a common lexicon and ontology for optimal functionality.
Of course, this information exchange must be bidirectional, with clinicians receiving timely and highly contextual data from the radiologist via pathways maximized for efficiency. Direct patient access to portions of the individual medical record is already in place in some institutions. Proper enterprise-wide integration will create efficiency and increased quality of care for radiologist, clinician, and patient.
- Communication. Many of us forget that the "C" in PACS stands for communication. We can use PACS as a communication tool in many ways. Communication with ourselves by teaching file or interesting case list creation is relatively simple. Communication with clinicians through receipt of clinical history and return of the radiology report is being performed at a rudimentary level by most PACS software, but there is much room for improvement.
Third-party systems are available that facilitate communication between radiologist and clinician. Whether third-party or native software is used, the communication system must be seamless on both ends. It must allow the radiologist to easily create and send annotated key images attached to an imaging report at a specified level of severity. It should allow the clinician to choose the method of receipt of these reports, be it telephone, pager, e-mail, or PDA, depending on assigned severity level as well as clinician schedule at any given time of day or night. It must also facilitate real-time communication, and it should include, of course, audit trail functionality.
- Decision support. Computer-assisted detection software has been in use for a number of years in mammography, lung nodule imaging, and CT colonography. New algorithms are being developed at a rapid pace, and this functionality is becoming the standard of care in many areas. CAD must be fully integrated into the reading process to maximize efficiency. Other current decision support tools include online access to textbooks, medical literature, and Web sites for image comparison and background information. PACS software should facilitate automated contextual access to the user's preferred method of information gathering. Anything short of this would be the digital equivalent of keeping a frequently referenced textbook on the other side of the department rather that at the workstation.
Other software being developed may benefit radiologists. Noise reduction filtering uses sophisticated mathematical algorithms to reprocess images. This technology has been used in the astrophysics industry for analysis of Hubble telescope images as well as in the military for satellite reconnaissance. It shows promise in the medical field for image enhancement in various modalities. Dr. James Wang's research group at Penn State University is developing an automated system for picture annotation using image recognition software. This process may ultimately prove helpful in data mining, image comparison, and teaching file creation.
- User training. The third requirement for disruptive technology is user expertise. Well-trained soldiers were necessary for the longbow's success. An English military archer of the 14th century could fire 10 to 15 arrows per minute and consistently hit a man-size target at 200 to 400 yards. Just as the English archers needed to train for the longbow, so must radiologists go through a learning curve in soft-copy interpretation. We have already largely converted to reading in stack mode. We must now train ourselves to read rapidly and accurately in coronal, sagittal, and nonorthogonal planes as well as in 3D and 4D. The interpretation and reporting process must also be linked in novel ways to allow report creation without distraction or time penalty.
All these methods for maximizing radiologist efficiency are laudable, but they do have limits. In electrical engineering, the exponential increase in the processing speed of semiconductor chips has created an unwanted side effect: increased power consumption. The technologic development of a lightweight, portable, and long-lasting power supply must keep pace with semiconductor advances. Engineers now measure performance/watt.
An analogy can be drawn to the radiologist's work day. When we arrive at work, our "batteries" are fully charged. A combination of time and task intensity level causes a gradual drain to near empty by day's end. Our stamina may diminish with age. Even with maximum PACS efficiency, the ultimate limit to image interpretation may be the intrinsic frailty of our own minds and bodies.
All designers must be aware that complex cognitive processes during image interpretation require full concentration and maximum eye dwell time on images. Future hardware and software design as well as workflow changes must be based strongly on these principles.
This article was adapted from a SCAR University presentation prepared by Dr. Weiss for the SCAR 2006 Annual Meeting. SCAR 2006 Syllabus, (c)2006 Society for Computer Applications in Radiology; www.scarnet.org.
Dr. Weiss is clinical section head of imaging informatics at Geisinger Medical Center in Danville, PA. He is a consultant to Agfa and a medical adviser to Commissure.