as long as the segmentation output is a raster, it is pretty unusable in SNAP for ongoning analyses.
However, having the segments as a vector which would then allow polygon-wise raster statistics could be interesting.