An example of the technic follows with the recognition of a face on a frontal extra-oral picture. A face is applied on the initial picture, and that drawing of the face is altered to match the facial characteristics of another person.
Initial grey part is of the face used as a "mold" to recognize and plot points at the face of this picture.
See how the initial face drawing, is at the end altered to trace the face of the new picture.
Similar approach is used in tracing a lateral cephalogram, using "mold" tracings that are applied to the X-Ray and then altered to match each patient's landmarks.
On the June issue of the EuJO, there is a paper that tests the reproducibility and accuracy of an automated tracing on 60 randomly selected lateral cephalograms.
- Accuracy of 1.68mm (mean value)
- 60% success of pointing landmarks 2mm apart from each-other
- 95% success on landmarks 5mm apart from each-other
Those findings are close to a well trained clinician, manually plotting the landmarks with a software.
The refining of the algorithm, and specialization to lateral cephalograms usage, could increase the accuracy in the future, giving accuracy higher than manual tracing, that of-course since it is the computer plotting the points, the reproducibility will be a lot higher than that of a human.
In the near future, clinicians will be able to have the cephalogram taken and have the analysis ready on the spot, enhancing the clinical examination of the patient and the treatment option discussion - evaluation the day of the registrations, at the dental chair.
Links of relevance:
http://ejo.oxfordjournals.org/cgi/reprint/32/3/233 (access needed for the full article)
http://www2.imm.dtu.dk/~aam/
http://personalpages.manchester.ac.uk/staff/timothy.f.cootes/
No comments:
Post a Comment