I am looking for various approaches for cancer segmentation and classification and was recently looking for Active Shape Model. The machine learning part of ASM is mainly based on Principal Component Analysis. I was looking for the feasibility of mixing ASM with deep learning? Can that be a feasible approach for segmentation pertaining to cancerous regions?
A part of an Active Shape Model is an
energy term that expresses the sum of the
distances of the points of the active contour
to their nearest point of the border of the
object. Somehow the border of the object
of interest must be identified. This might
be possible with and edge detector, but
even better but slower would be to use
a neural net that can classify borders.
In a further step you can apply a
Gauß filter on the edge image.
Then you get an image that expresses
how near you are to a border.
Regards,
Joachim
Thanks Abdul,
ASM has its own merits and demerits when applied to the particular problem. It has to be hybridized with some techniques , so was exploring Neural Nets in order to get more optimal results.
- Badges
- Science topic