| |
|
|
|
|
|
|
Data Analysis
Back to Data Analysis
3. Chronic Wounds
To characterize the wound healing process, we conceive a wound image as a
set of elements that can be hierarchically indexed. For instance, we define
the top level as ''Skin'' and the second level separates ''Healthy Skin''
from ''Wound Area''. In this way, it is possible to define deeper levels
like ''Wound Border'', ''Wound Core'', Fibrin, etc.
Using a variety of methods, we extract several image characteristics for the
different elements. We consider many color spaces and also methods that account
for the structural content of the image as for example the Scaling Index Method.
With the help of an information theoretical analysis, we are able to sort out
image features and select those that are complementary. The unsupervised
segmentation of the different elements for new images is achieved through a
classification algorithm that considers ''memory based reasoning''. Our results
show an average classification rate of approximately 0.82%.
Read more:
- Data Mining Aspects
Top of page
|
|
|
|
| |
|
|
