“…Due to the incompleteness of this information, current commercially available tracking systems (e.g., superDimension), which are increasingly used in clinical applications, [9][10][11] reported their diagnostic accuracy between 59% and 74%, without sensitivity to tumor size. 12,13 To tackle this incomplete information for accurate endoscopic guidance, it is common to formulate such multimodal information fusion as an optimization process, which is usually solved by deterministic 14,15 or stochastic [16][17][18] approaches. Deterministic methods, typically 2D/3D image registration algorithms, 4 usually define an optimization function to minimize the pixel difference between endoscopic video images and virtual renderings generated from preoperative information.…”