surface sliding semilandmarks in patches

[Alexandra Barnard]

I want to quantify a plate-like structure on an insect thorax, and I have 3D mesh reconstructions from CT scans.  The plate is not easily segmented from the rest of the thorax, so I was hoping to place landmarks just on the plate and ignore the rest of the thorax.  Is there a way to assign surface sliding semilandmarks to a specific region of a 3D mesh rather than placing them evenly all over the surface?  

I have assigned landmarks using Avizo and read their coordinates into Geomorph.  I tried making a template.txt file from one specimen’s landmarks (18 fixed plus 6 surface sliders) and then using digitsurface() to apply those landmarks to other surfaces, but the landmarks end up clumped together and in the wrong location - even on the surface from which I made the template.

I welcome any suggestions.

[Emma Sherratt]

Dear Alexandra,

geomorph does not currently have a "buildpatchtemplate" function. The process of defining and placing semilandmarks to slide over a 3D surface in a demarcated area ("patch") is not trivial and differs substantially between objects for feasibility, but a few options exist. 

IDAV landmark (and it's relative Checkpoint) allow a user to place a patch, defined by a set of fixed landmarks around the perimeter, wherein equally-spaced semilandmarks landmarks arranged in a grid are placed on the 3D surface. In this program, the user must place a patch on every specimen, and the semilandmarks are calculated as equally spaced each time. This can mean than for wildly different patch shapes, the semilandmarks are not strictly positionally homologous. 

An alternative that is growing in popularity is to build a patch template (using a specimen that is rather "average") in IDAV landmark or other software, and then fit this patch to each specimen in R:  specifically taking the template (which is a grid of semilandmarks, and a perimeter of fixed landmarks), warping it using thin-plate spline based on the fixed landmarks on a second specimen, and setting down the positionally homologous semilandmarks on each 3d surface. One could do this with geomorph's digitsurface or morpho's placepatch. This approach requires the fixed landmarks to be digitised elsewhere (using digit.fixed in R, or in IDAV landmark, or AMRIA etc.).

There also exists Viewbox which fits a patch/template using TPS within the digitizing software itself, thus eliminating steps where one has to digitize fixed points in one software, and fit teh patch in another.

Regardless of the way these patches are placed on the 3D surfaces, the sliding is done during Procrustes superimposition.

I look forward to users chiming in here with their experiences and workflows using 3D semilandmarks.

Emma