3D intra-operative scene reconstruction framework in robot-assisted minimally invasive surgery

The introduction of robotics in minimally invasive surgery overcomes many obstacles compared with traditional open surgery. It also opens the way for the integration of artificial intelligence in surgery. Image guidance, registering pre-operative model photoed through computed tomography or magnetic resonance imaging into intra-operative soft tissue, has been reported to improve the safety of surgery. The performance of intra-operative scene reconstruction directly affects the reliability of this image guidance technology. Therefore, we propose to design a computer vision-based architecture to reconstruct the 3D appearance of the surgical scene using stereo image pairs from an endoscope.