Event Location: NSH 1507

Abstract: In the push to improve patient outcomes in cardiac interventions, minimally invasive beating-heart surgery is a major field of surgical research. However, interventions on a soft tissue organ under continuous motion through remote incisions pose a significant challenge. Endoscopic approaches eliminate the associated morbidity of median sternotomy, but they require either mechanical immobilization of the heart or robotic motion compensation of the tools, both of which have serious drawbacks. While mechanical immobilization may cause electrophysiological and hemodynamic changes in the performance of the heart, active compensation requires high-bandwidth manipulators to track the complex motion of the heart.

In this thesis, we address the issue of physiological motion during minimally invasive beating-heart surgery through the use of organ-mounted robots. These devices eschew the high dexterity and actuation effort required of traditional surgical robots in favor of miniature robots that adhere directly to the operating site using vacuum pressure. Unlike mechanical stabilizers these devices are not fixed in the world frame and therefore do not immobilize the heart but instead move in unison with the heart providing a stable platform from which interventions may be administered.

This thesis is built around two main contributions to the state of the art in robotic MIS. First, we design, build and demonstrate new robots that provide a wider array of interventions using the organ-mounted paradigm. These efforts use emerging therapies as motivation for the design of an active cooling system for minimally invasive delivery of thermosensitive materials and a new parallel wire robot, known as Cerberus, for therapies that require coverage over large areas of the surface of the heart. Both of these new capabilities are demonstrated successfully in closed-chest beating-heart procedures.

The second major contribution of this work is the development of spatiotemporal registration methods to improve positioning accuracy under virtual image guidance for organ-mounted robots. These efforts rely on frequency-based models, which capture the periodic motion of the heart, and anatomical models constructed from preoperative imaging. Using these models we estimate when in the physiological cycles the images were captured and the pose of the robot at that time to spatially align the models. Finally, we introduce a method for localizing these robots on the beating heart using function approximation that provides more accurate estimates over short time horizons.

Overall, our contributions take a holistic approach to the advancement of the capabilities of organ-mounted robots. New form-factors provide specialized capabilities, while new approaches to registration improve our ability to accurately position these robots on the beating heart. Most importantly, everything presented in this thesis is demonstrated in closed-chest beating-heart procedures, or on data recorded in such a procedure.

Committee:Cameron Riviere, Chair

Howie Choset

George Kantor

Pierre Dupont, Harvard Medical School