If you want to make a more dilated volumetric rendition of a neuron, you can either
* increase the voxel size (by making an object with the same bounding box as JRC2018U) but larger voxels
* do some appropriate 3D convolution on the line object.
The natverse is not particularly set up for the latter, but I have used Fiji/ImageJ for this purpose after writing out a nrrd file as you have done below. You could use e.g. the Gaussian 3D operation with a 2 µm kernel.
When I had a lot of this work to do, I have also used the unu command line tool [1] resample command to do the filtering. There are a a range of alternatives to the gaussian kernel which may be better in retaining the sparse nature of the dataset. One other thing, it is important to save compressed nrrds (e.g. gz) for this kind of work since the image volumes will only have a few non-zero pixels.
All the best,
Greg.