The introduction of Robotics in Minimally Invasive Surgery (RMIS) allows overcoming many of the obstacles introduced by traditional laparoscopic techniques, by improving the surgeon’s maneuverability and precision during the surgical procedure. Nevertheless, major complications can still affect the outcome of the procedure, for example intra-operative bleeding caused by vessel injuries.
The overall goal of this thesis is to develop novel computer vision algorithms for the implementation of assistive technologies focused at enhancing the surgeon’s intra-operative capabilities, to allow safer clinical procedures and improved outcomes.
The PhD research has been focused on the following topics:
(I) A dense 3D reconstruction algorithm was developed to intra-operatively measure soft tissue deformations. The work was focused on the refinement of the disparity map with the aim of obtaining an accurate and dense point map. In order to evaluate the algorithm, a phantom of abdomen (liver, kidneys and spleen) was developed and a new and rich stereo endoscopic image dataset (EndoAbS dataset) was created.
(II) A framework for 2D long term tracking of soft tissue areas, to be preserved from injury during surgery, was developed. A novel strategy to recover failures from tracking was proposed to make the algorithm robust against: (i) partial occlusion, (ii) total occlusion, and (iii) camera movements.
(III) Development of an Enhanced Vision System to Improve Safety in Robotic Surgery (EnViSoRS) to protect vessels from injury during the execution of a robotic surgical procedures. The core of the framework consists in the integration of the previously developed dense 3D reconstruction algorithm and SA tracking, and Augmented Reality (AR) features for warning the surgeon about distance between the instruments and the vessel.