When it comes to medical purposes, hybrid approaches, that combine 2D picture options with depth information to reconstruct 3D objects, have been mostly used. No ML models are excellent and we should admit that our model would possibly sometimes produce unexpected predictions