The use of voice recognition software to trigger commonly used actions on the PACS workstation will reduce the amount of distraction and input device switching during the reading process. This will reduce the time needed to read studies and result in a more satisfying reading experience.

The challenge of radiology workstation user interface design has always been to minimize the intrusion the software itself imposes on the reading process; in other words, to help radiologist focus on reading studies by minimizing any interruption to the diagnostic train of thought or visual fixation on the images. The common paradigms of triggering system functionality using menus, toolbars, and keyboard shortcuts interrupt the reading process in both ways. First, radiologists are required to think about how to invoke the functionality they desire (i.e., toolbar, menu or keyboard shortcut). Second, they are required to take their eyes off the image (e.g., to navigate and select from a menu). By allowing radiologists to speak commonly used commands, mental interruption is minimized, and interruption of visual fixation on the images is avoided.

In addition, radiologists typically have 3 input devices: keyboard, mouse, and speech microphone. Having only 2 hands and switching between these devices is a necessity, especially when the radiologist repeatedly switches between navigating/manipulating the images and dictating. Again, by allowing the radiologist to speak commonly used commands, the amount of switching between input devices is minimized.

It is interesting to note that extended interaction with the mouse and keyboard introduces the risk of repetitive strain injury. While this abstract does not seek to measure the benefit, it seems intuitive that any reduction in mouse/keyboard use will reduce the risk of repetitive strain injury commensurately. Voice control should also be of great interest to radiologists already suffering a repetitive strain injury.

Voice Recognition software was installed on a PACS workstation and voice commands were programmed which triggered commonly used mouse and keyboard shortcuts. These included:
-Study navigation: next/previous page, next/previous display protocol stage
-Image manipulation: zoom, flip, rotate
-Image processing: invert, window/level presets, magnifying glass on/off
-Other: select annotation tool, turn overlays on/off, show study documents, close study

Features not included were those involving interactive mouse movement, such as power scrolling, and interactive window/leveling and panning. Early prototypes revealed that interactive functions are difficult for users to manage using voice and thus are counterproductive.

The hypothesis was tested by timing a group of radiologists reading a set of studies both with and without voice control on the reading station. Participants were divided into 2 groups to examine variability between the types of users: “Regular Users” and “Power Users.” Regular Users are defined as radiologists with average computer experience who primarily use the workstation “out of the box” and have made minimal effort to tune or customize the workstation to suit their reading preferences. By contrast, Power Users have above average computer savvy and have their workstation shortcuts and preferences highly tuned to suit their preferences and the types of reading they perform.

Because this abstract is attempting to measure the improvement voice control can offer the reading process (i.e., not the effectiveness of the voice control technology itself) several constraints were put on the study to control for factors known to confound voice recognition:
-Participants with heavy accents were excluded
-Ambient noise in the reading area was controlled
-Participants went through the recommended standard training procedure with the voice recognition package
-Participants were given sufficient time to learn and become comfortable using the voice commands

This study is currently in process. It is expected that the results will show a small, but measurable, decrease in the amount of time required to read a study. The effect is expected to be more pronounced for typical users of PACS, while power users realize a lesser impact.

Because of a reduction in mental and visual distraction, adding voice control capability to a PACS workstation will allow radiologists to read faster and to have a more satisfying reading experience.

For further study, it would be interesting to determine how a study group with existing repetitive strain injury (RSI) would perform. It is expected the increase in performance would be larger for this group when compared to non-RSI groups.

Furthermore, a study of the impact a voice control system would have on the learning curve for new PACS users would be interesting. Because voice commands make use of natural language, they have a distinct advantage over keyboard short cuts. For example it is much easier to remember to say “zoom” than to remember that “Ctrl-z” is the keyboard shortcut. This effect should be more pronounced as the number of commands needed to be remembered increases. Having a menu to choose from acts as a reminder of what commands are available. However, this problem is easily avoided by keeping the menu option available and by keeping voice commands congruent with the captions on the menus.