Optimization-Based Design of a Robotic Endoscope for Intraventricular Tumor Surgery

Tumors in the ventricular system can be surgically removed via neuroendoscopy. This minimally invasive procedure involves introducing an endoscope through a small incision in the skull until reaching the tumor’s location. In these surgeries, there are places where a rigid endoscope has difficulty reaching. Therefore, it forces the surgeon to use two incision points. Flexible endoscopes facilitate reaching challenging regions with only one incision, and a flexible robotic endoscope helps surgeons orient and navigate inside the patient. Although the flexible endoscope is convenient to use, there is no guideline on how its structure should be composed. In this thesis, we used an articulated flexible endoscope, and the question about its structure is: how long should the segment be? How many segments should the endoscope have? and how long should the total length be? The scenario for this thesis is a neuroendoscopy in an area where a flexible endoscope is convenient since it is difficult to reach the posterior third ventricle. We proposed to optimize the endoscope’s structure based on the final configuration reaching the posterior third ventricle. The hypothesis underlying this project is the following: Can the structural parameters of a robotic endoscope be optimized based on the desired endoscope’s end configuration for the tumor removal in the posterior third ventricle? The proposed framework uses a genetic algorithm to optimize the area between the desired path and the final configuration of the endoscope, and the distance outside of the path. The cost function was derived from an inverse kinematics solver. We first used Matlab’s [1] global optimization toolbox to create the genetic algorithm and robotic system toolbox to create the inverse kinematic solver. When the first framework showed results, further steps were made to create a more elaborate framework with ROS[2], Move it, and OMPL[3]. Where we kept the genetic algorithm but changed the inverse kinematic solver to a motion planner, to answer the next steps of optimizing the structure of an endoscope, ”Find an optimal design of an articulated flexible endoscope based on the motion to reach the tumor in the posterior third ventricle. ” The new cost function is achievable goals and follow the leader, and both of them are calculated when the motion planner is active. Although this approach has the potential to give better results than the first approach, it still needs more work to be able to return results.
The ability of the first framework to optimize the length of the endoscopes’ segments was quantified under three configurations with different segments available four, five, and six segments. We saw that the various configurations had their optimized structure relative to the desired path, how they evolved, and the tendency that leads to a low cost structure. It also displayed their cost values through each generation. Although with some limiting factors, the results show that we could answer the question made for this thesis. The results showed low cost structures that tried to be close to the desired path and reduced the length that was not used on the inverse kinematics solver. These new frameworks need more work, but they open a path to future students or researchers who will continue this research area of optimization of endoscope structure.