Med. Collab. Prototype Test Results.
Report by: Beth, Tom, and Stephane.
The Prototype test occurred last Tuesday morning (2/19/96). Those present
included Ron, Tom, Todd, Beth, Stephane, and Nelson.
The test went well. Ron was able to annotate images and was generally pleased
with image quality. However, the test revealed a number of areas for improvement.
These are outlined below in order of their importance.
1. Pointing
In consultation situations, pointing serves to establish common ground.
Radiologists will point to an image to highlight a feature for the clinician
or demonstrate a finding.
During the prototype test, Ron tried to create some pointing metaphor.
Since the system could not synchronize pointer movements and window location,
Ron tried anatomical labeling, which was very awkward and time consuming.
After that he tried the region of interest strategy, but concluded that
pointing was the most desirable.
Recommendation: Pointing is a necessary feature for support of the consultation
process.
2. Emulate Dictation
The dictation process is a amalgamation of the interpretation and dictation
process for attending radiologists. Therefore, when dictating radiologists
often stop, start, and revise, and play back their dictations.
During the prototype test, Ron unconsciously does this and this causes problems
for the final product because the system does not support this practice.
Recommendation: Record, pause, playback, and edit are necessary features
for support of the dictation process.
3. File management
Under traditional circumstances images are presented for the radiologist
and the radiologist must interpret and read them out.
During the prototype test, Ron had to find his own files, load them into
the Osiris system, which required him to know the exact name of the files
in order to type them in. This also required him to know and remember the
file contents. Therefore, much of RonUs time was occupied trying to figure
out which file to look at and getting it into the system.
Recommendation: The user needs a convenient file management system which
does not require memorizing the contents of the files or typing in complicated
file names.
4. Window Management
Radiologists view images as components of larger studies and often need
to compare these with prior groups of images (other studies). In this way,
radiologists get the RgestaltS of the case (e.g., Radiology the Movie, the
first scene with Bowerman).
During the prototype, Ron tried to engineer a mechanism for viewing multiple
images. This worked from the dictatorUs point of view, but on playback
lost context, (e.g., the order of the image selection was not preserved),
such that dictation was not synchronized with the appropriate image.
Recommendation: User should have a means of associating images together
as studies or cases. Dictation must also be synchronized with the correct
images.
5. Image Annotation
Radiologists rarely mark up images. During the prototype test, Ron tried
two strategies. First, he labeled anatomical features. Second, he identified
regions of interest and talked about them. To a certain extent, this was
artifactual because Ron was trying to respond to the absence of pointer
recording. To save time, Ron wanted to annotate images in advance and then
dictate them. The system would not allow him to save annotated images.
Recommendations: Solve the pointer problem. Create a way for the radiologist
to save annotated images.
6. Language
Radiologists do not use "region of interest" as a term of art.
During the prototype, Ron was not comfortable with this terminology.
Recommendation: Don't impose jargon.
7. Information Overload
In clinical practice, clinicians review images in 30 seconds or less (e.g.,
MedSport). During the prototype test, the system recorded all of Ron's
actions, most of which are overhead (including extremely awkward start-up
for the Osiris viewing system) related to Ron selecting and annotating images.
Recommendation: See above about revising and editing.
8. Specificity of Tools
Each tools is loaded with several different functions. For example, the
region of interest tool always comes up with an arrow. During the prototype
test, sometimes Ron only wanted the arrow.
Recommendation: Decouple tools, for example just give Ron an arrow tool.