We’re interested in understanding how VTK can facilitate the conversion of RTStruct contour data, which includes floating-point coordinates, into 3D models. Specifically, we have several key questions:

Does VTK offer a process for directly converting RTStruct data with floating-point contour coordinates into 3D models? We understand that conventional algorithms like the marching cubes typically operate on integer-indexed 3D grids.

In our attempts to convert RTStruct contour data with floating-point coordinates into an image representation, we observed that rounding to integers (as each pixel can only be either 255 or 0) alters the resulting STL model when generated using the marching cubes algorithm. Is there a VTK-specific method that mitigates this rounding issue, preserving the original shape dictated by the floating-point contours? As an additional note, we’ve tried image scaling by a factor of 10 as an initial approach, but found it to be computationally intensive. For example, scaling a 512x512 image to 5120x5120 for just one slice becomes problematic when working with a minimum of 160 slices in our use case.

If VTK can operate directly on floating-point contour data, what techniques or methods are utilized to fill the surface between inner and outer contours, thereby creating a complete 3D model?