search

SCIInstitute / Seg3D

Seg3D is a free volume segmentation and processing tool developed by the NIH Center for Integrative Biomedical Computing at the University of Utah Scientific Computing and Imaging (SCI) Institute.
108 stars 47 forks source link

different interpolation method with the polyline tool #239

Open jessdtate opened 7 years ago

jessdtate commented 7 years ago

We get intermittent requests for cubic and other interpolation schemes with the polyline too. This is something that a third party implemented in Seg3D 1, and I think it is in corview. This is the text of the latest request.

I would like to recommend adding an option/function/button with the polyline that allows you to create a bicubic interpolation between the points of the polyline. For example, a user would place something like 15-20 points along a curved border (which would normally create a relatively linear-looking ROI) and then the user would click on a bicubic interpolation button. The polygon would then become a nice, smooth bicubic interpolation creating a better approximation of a given curved border and saving considerable time in the process. The user would then fill-in the curved line to create their new, more accurate mask layer. I would also recommend a 'revert' feature that will revert back to the original linear polyline to allow for adjustments, such as re-positioning points or adding points to better deal with very tight curves.

jessdtate commented 6 years ago

From Marcus Bobar, a Seg3D user.

Recently, I've developed a couple different algorithms in MATLAB using a piece-wise periodic spline curve interpolation (link and some info below) with great results and I think that it could be applicable to a potential algorithm in Seg3D. You just need to have the points in order and the first and last points be the same. I assume the language in which Seg3D is coded is different than MATLAB, but it seems like the algorithm should be adaptable without too much trouble. What I've done once I have the curve and the break points, is express the piece-wise function evenly around the curve for a number of points and then I used an 'in-polygon' function (link below) to create a mask for the pixel points that lie inside the curve. I'm an ok programmer at best and even with my average skills the algorithm was relatively quick to run for a given ROI.

I'm passing along this info because: A. Our lab could really, really use interpolation to create curved borders for cardiac MR images more quickly; B. I'm sure you are always busy just like I am, so I figure some preliminary work being done is helpful; C. We would love you forever if you can incorporate interpolation and I promise to mention you in the next couple papers we right this year for your efforts. Thank you in advance for your consideration.

Spline Curve:

https://www.mathworks.com/help/curvefit/cscvn.html

"curve = cscvn(points) returns a parametric variational, or natural, cubic spline curve (in ppform) passing through the given sequence points(:j), j = 1:end. The parameter value t(j) for the jth point is chosen by Eugene Lee's [1] centripetal scheme, i.e., as accumulated square root of chord length.

If the first and last point coincide (and there are no other repeated points), then a periodic cubic spline curve is constructed. However, double points result in corners."

In-polygon:

https://www.mathworks.com/matlabcentral/fileexchange/10391-fast-points-in-polygon-test