This summer at Dexterous Endoscopes, an EPFL spin-off, I built a testing ground for a nephroscope used in kidney stone removal: a silicone kidney phantom derived from a real patient's CT scan, reproducing the internal collecting system the device needs to navigate rather than just the kidney's outer shape. The pipeline covered isolating the collecting system from CT data in Blender, designing a two-part mold with alignment pins and a syringe injection port, and casting the geometry in silicone.

A phantom is only useful as a testbed if it's actually faithful to the anatomy it's meant to represent, so I validated the cast: scanning it by photogrammetry and comparing the resulting mesh directly to the source CT model, giving a mean deviation of ~0 mm (SD ~0.52 mm). With that quantified, I used the phantom to test the nephroscope's ability to navigate and access three renal calyces from separate percutaneous entry points, evaluating it as an early, low-cost testbed for device performance ahead of clinical trials.