For most use cases, the 1.5mm resolution of the segmentation is good enough. You often have to work with lower-resolution input images anyway. Discussion about improving the resolution only makes sense if you have a specific goal in mind, as creating perfect segmentations (that are suitable for every possible applications) is impossible. What is your clinical application?
Note that if the goal is to just to create nicer visualization then surface rendering is probably not a good approach anyway. Segmentation-enhanced volume rendering has much bigger potential for visualization - see for example in this topic.