(Georgia Institute of Technology, 2010-06)
Hoenig, Helen; Griffiths, Patricia; Harris, Frances; Caves, Kevin; Sprigle, Stephen
This study examined
the accuracy of
new wheelchair user predictions about their future
wheelchair use.
We used an existing
database of 71
new manual wheelchair
users with data
obtained at
baseline, 3-‐ and 6-‐months to examine the
specificity, sensitivity,
positive and
negative predictive
value of user
predictions
about anticipated
amount and
locations of
wheelchair use.
At 3-‐months,
the correlation
between predicted
and actual use was
strong, with
90% of those
who thought they
would still be
using the wheelchair
still using it,
and 60% of
those who said
they would not
be using it
indeed were not
using the wheelchair.
By 6-‐months the
predictive utility
diminished substantially.
Only 70%
of subjects
accurately
predicted
their continued
use, while
only 50% correctly
predicted they
would not
be using their
wheelchairs.
This study
demonstrates
the importance
of better
understanding
the potential
mismatch between
the anticipated
and actual patterns
of wheelchairs use.
(Georgia Institute of Technology, 2006)
Sprigle, Stephen; Patel, Nirmal; Joshi, Aditya; Starner, Thad
Repeatable and accurate wound measurement forms an important part in the assessment
and treatment of chronic wounds and pressure ulcers. Current wound measurement methods span
a continuum, from the ruler method which is easy to perform but lacks accuracy to devices using
stereophotogrammetry which are accurate and repeatable but are expensive.
A prototype handheld wound measurement system has been developed that measures
wounds without contact, processes images in <1 min and is low cost. The device is based upon a
simple digital camera such as those found in cell phones. Using computer vision techniques,
device software suggests a wound boundary and gives the calculated area. The user can then 1)
accept the area (if the wound boundary detection is correct), 2) modify the wound boundary by
dragging the outline using a stylus on the touch screen, or 3) reject the wound boundary and retrace
the wound manually using the stylus.
Accurate wound dimensions using photography require knowing the distance between the
camera and wound and the ability to correct for a skewed image. The use of four laser pointers
and computer vision techniques overcome these technical challenges. To test accuracy,
repeatability and skew correction, a model wound with known dimensions was measured at four
distances and skew angles between 0º and 35º. Accuracy across distance and skewness ranged
from 5%-7.5% with a coefficient of variation (repeatability) of <4%. This performance exceeds
values reported for rulers, tracing methods and photography and equals performance of higher
cost structured light devices.