The original developers of VMTK used it mainly for vascular applications, but it is equally applicable for analysis of any somewhat elongated, tubular-like structures (has been used successfully for airways, intestines, plant roots, etc.). It should work well for sutures, too.
Yes, you are right, you won’t be able to see cube voxels. It is because segmentation is for representing smooth, continuous surfaces. Even if in some representations (such as in binary labelmap) discrete point samples are used, the result is always reconstructed into a smooth surface. For binary labelmap to closed surface conversion we use continuous version of the flying edges algorithm, which creates rhomboids from standalone voxels (and not the discrete version, which creates cubes).
You could run the discrete version manually by using a couple of lines of Python script, but I would not recommend that, because physical reality is never discrete (at least not in our domain). Instead, properly segment the real size of the suture, and then get the smooth continuous centerline curve directly from that, using VMTK.
If you don’t have the original segmentation anymore, just the binary labelmap skeleton then you can use Margin effect in Segment Editor to blow it up into a thicker tube. You can then use VMTK on this thicker segment.