«Dexterity optimization by port placement in robot-assisted minimally invasive surgery Shaun Selha1, Pierre Dupont1, Robert Howe2, David Torchiana3 1 ...»
2001 SPIE International Symposium on Intelligent Systems and Advanced Manufacturing, Newton, MA, 28-31 October
Dexterity optimization by port placement in robot-assisted
minimally invasive surgery
Shaun Selha1, Pierre Dupont1, Robert Howe2, David Torchiana3
Aerospace and Mechanical Engineering, Boston University, Boston MA, USA;
Division of Engineering and Applied Science, Harvard University, Cambridge MA, USA;
Division of Cardiac Surgery, Massachusetts General Hospital, Boston MA, USA
ABSTRACTA computer-based algorithm has been developed which uses preoperative images to provide a surgeon with a list of feasible port triplets ranked according to tool dexterity and endoscopic view quality at each surgical site involved in a procedure. A computer simulation allows the surgeon to select from among the proposed port locations. The procedure selected for the development of the system consists of a coronary artery bypass graft (CABG). In this procedure, the interior mammary artery (IMA) is mobilized from the interior chest wall, and one end is attached to the coronary arteries to provide a new blood supply for the heart. Approximately 10-20 cm is dissected free, using blunt dissection and a harmonic scalpel or electrocautery. At present, the port placement system is being evaluated in clinical trials.
Keywords: Port Placement, Dexterity, Minimally Invasive Surgery, Telesurgery.
1. INTRODUCTION Robotic assistance enables the use of minimally invasive techniques in coronary artery bypass grafting by scaling hand motions, decreasing tremor, and enhancing manipulation. In these procedures, endoscopes and other instruments are inserted through small incisions or “ports” to access the surgical site. In initial trials of this approach, inappropriate port locations have posed a number of difficulties including poor tool dexterity, an inability to reach all the required surgical sites, and an increased likelihood of collision between the instruments, both within and outside the patient’s body.
This paper presents an algorithm for optimizing port placement for the ZEUSTM surgical robot system. The three main issues in defining optimal port placement are tool dexterity, endoscopic viewpoint, and workspace limits. Tool dexterity is defined by the relative orientation of the tools with respect to each other and the surgical site. Endoscopic viewpoint is defined with respect to the tools and with respect to the surgical site. Workspace limits are reached when a desired tool motion would cause a tool/ endoscope/ patient collision (internal) or a robot/ robot/ patient collision (external).
2. BACKGROUND Several medical research groups have addressed the problem of optimal port placement for robotically assisted minimally invasive surgery, with respect to tool dexterity, endoscopic viewpoint, and workspace limits1-3. Coste-Maniere et al. have presented a pre-operative planning and simulation system for minimally invasive surgery using the daVinciTM robot system1,2. The approach consisted of a planning, validation and simulation phase. Criteria considered in the port location process included the surgical target site, the attack angle of the instrument, describing the angle between a normal to the surgical site and the instrument, and a dexterity parameter, describing the angle between the surface normal of the port and the instrument. Optimal angles were based on the experience of a cardiac surgeon. In this approach, possible port locations for two instruments and an endoscope are ranked according to these criteria as well as the potential for external robot collisions.
Contact information: Shaun Selha (firstname.lastname@example.org), Pierre Dupont (Pierre@bu.edu), Robert Howe (email@example.com), and David Torchiana (firstname.lastname@example.org) Experimental guidelines for the selection of port locations for the ZEUSTM robot system made by Computer Motion, Inc.
are presented by H. Tabaie et al.3. Based on dry lab experiments, a template was developed indicating optimal robot configurations as well as ideal instrument and endoscope port locations. A port template diagram defines the optimal relative locations and orientations of the right and left instruments and the endoscope. A similar diagram is employed here and will be presented in section 3.
Within the robotics literature, the issues of tool dexterity, endoscopic viewpoint, and workspace limits can be related to robot dexterity4-10. Many of the reported dexterity measures have been proposed to measure endpoint dexterity for multilink robots. Consequently they involve some function of the singular values of the Jacobian matrix, thus penalizing robot arm configurations for which end point translations and rotations require large joint motions. The most referenced of these is the manipulability measure proposed by Yoshikawa4, which is defined as the square root of the determinant of the matrix JJ T. This is simply the product of the singular values of J. Other proposed measures include the minimum singular value of the Jacobian, which was proposed by Klein and Blaho5, as well as the trace of the matrix JJ T, put forth by Baillieul6. Other measures include the compatibility index formulated by Chiu7, and isotropy (all equal singular values) proposed by Salisbury and Craig8.
An alternate approach to defining dexterity is the joint range availability function first proposed by Liegeois9. In this approach, dexterity is highest when joints are centered in specified ranges. This approach has been recently augmented by Liang and Liu10 such that the desired midrange angle of each joint is switched at a series of pre-specified path points of the end effector. They also proposed applying variable weightings to the angles in order to achieve the proper postures effectively.
In the next section, the port placement problem is defined and the appropriate approach to measuring dexterity is explained. In the following section, an overall dexterity metric incorporating tool dexterity, endoscopic viewpoint and workspace limits is formally defined, which is related to joint range availability. Implementation of the algorithm is detailed in section 5 and the last sections describe ongoing clinical trials and assessment goals.
3. PROBLEM DEFINITION – PORT PLACEMENT
In the minimally invasive surgical procedures considered here, three thin rods are inserted into the patient’s thorax through separate incisions in the chest wall, called ports. These incisions are made in the intercostal spaces, i.e., between the ribs. Two of these rods constitute instruments and terminate, interior to the patient, with tools, such as a gripper, scalpel or cauterizing tool. The third rod is an endoscope, providing a camera view of the surgical site. In the Zeus™ system, there is no articulation at the distal ends of the rods where the tools attach. There are often, however, fixed offset angles between a rod’s axis and its tool or endoscope.
The proximal end of the rod is attached, exterior to the patient, to a six degree of freedom robot with four active joints and two passive joints. The ports act as fulcrums for the rods. Motions of a tool or endoscope are accomplished solely by translations of the rod along its axis and rotations of the rod about its port.
The robots are clamped to the operating table and include two additional locking joints. By adjusting the clamping position of the robot relative to the surgical table and by adjusting the two locking joints, the robot can be configured relative to the surgical table and patient.
Dexterity-related difficulties during surgery can be attributed to the following factors:
1. Inability to reach the surgical site,
2. Inability to perform the surgical procedure due to the orientations of the tools with respect to each other and the surgical site,
3. Instrument/endoscope collisions occurring inside the body,
4. Robot singularities and joint limits, and
5. Robot collisions.
The approach taken in this paper is to decompose the overall dexterity problem into two sub-problems: port placement and robot placement. Port placement deals with difficulties 1-3 while robot placement controls difficulties 4-5. By employing this decomposition, the problem of port placement can be addressed independently in this paper. The port locations so obtained can be combined with a robot positioning solution based on either experience or an algorithm developed for the purpose.
By decoupling the problems of port and robot placement, the overall dexterity problem is also decoupled. Consideration of difficulties 1-5 above indicates that different dexterity measures may be appropriate to each sub-problem. Only port placement dexterity is considered here. The metric described in the following sections was developed through surgical observation and interview.
3.1 Port placement coordinates
The coordinate frame and relative angles used to define port placement dexterity are shown in Figure 1. For clarity, only the cylindrical rods of the instruments and endoscope are depicted along with a small planar patch representing the surgical site. A coordinate frame is defined relative to the surgical site such that the y -axis is normal to the site. For the procedures described here, the normal is always considered to be oriented vertically. In this case, the axes can be further specified such that the x -axis is taken to be parallel to the patient’s spine, and, together with the z -axis, lies in a plane parallel to the operating table.
The angles used to define the positions of the right and left instruments and the endoscope, each held by their respective robotic arms, can be described relative to this coordinate frame as shown in Figure 1. The two instruments lie in a plane (the instrument plane) with angle of elevation γ from the z -axis. Within the tool plane, the left and right instruments are oriented at angles θl and θr with respect to the negative and positive x -axis, respectively.
Endoscope orientation is described by three angles, φe, φa, φo. The first two, φe and φa, are elevation and azimuthal angles of the endoscope, respectively. Not depicted, φo is a constant offset angle between the endoscope’s axis and its line of sight. In the procedures under consideration, a 30-degree offset endoscope is typically used.
A qualitative explanation of the optimal angles can be given as follows. The relative orientation of the instruments and endoscope in Figure 1 is motivated by the desire to make robotic surgery intuitive. This is accomplished by preserving the relative orientations of the surgeon’s hands and eyes with respect to the surgical site, as displayed on a video screen.
Thus, the left and right instruments are constrained to share the same optimal angle of elevation, γ r = γ l. The optimal values are due to the relative anatomical locations of the IMA and heart since the same ports must be used for IMA mobilization (takedown) and anastomosis.
Intuitiveness also constrains the endoscope to be positioned between the left and right instruments. Furthermore, experience has shown that minimizing φa (i.e., φa,opt = 0 ) minimizes the length of the endoscope inserted into the patient’s thorax. This in turn minimizes the probability of an internal collision between the endoscope and the right and left instruments.
The endoscope elevation angles, not including offset, are 7 degrees for IMA takedown and 52 degrees for anastomosis.
In viewing the IMA, which runs along the underside of the sternum, the 7 degree elevation angle combines with the 30 degree offset to look upward at an angle of 22 degrees. During takedown, the instruments are also angled upward and must move through a large range of travel to the left and right. Achieving a viewing angle steeper than 22 degrees would necessitate insertion of a greater length of endoscope, which would generate collisions with the instruments.
For anastomosis, the 52 degree elevation angle combines with the 30 degree offset to achieve an 82 degree viewing angle. While an optimal viewing angle of 90 degrees could be achieved, the 8 degree offset makes it possible to flush the endoscope lens from an assist port during the procedure.
Among the orientation angles, experience dictates that the order of importance is θr and θl, followed by γr, γl, and φe, and then by φa. This indicates that the position of the surgical instruments is more critical than the position of the endoscope. The values of the weighting factors in Table 1 were chosen accordingly. Note that the optimal angles and weighting factors can be modified for other procedures or the preferences of other surgeons.
5.1 Algorithm A search algorithm and graphical user interface have been developed using the dexterity function of (1). The inputs and outputs of the system are listed in Table 2. For the planning to be based on a specific patient’s anatomy, a preoperative CT scan is required with a contrast agent used for artery imaging. The image must be segmented to obtain 3D models of the skin, rib cage, IMA and heart. Additional inputs include the type of procedure, a set of targeted surgical sites and a grid of feasible port locations.
The user selects a set of surgical sites and a grid of feasible port locations using a computer mouse to interact with the 3D models through a graphical interface. The grid of port locations is selected using the skin model and must be located in the intercostal spaces, i.e., between the ribs. For IMA takedown, four target sites, spanning the length of artery to be harvested, are specified. For anastomosis, a single target location must be specified for the coronary artery to be bypassed. As shown in Figure 1, surgical sites are defined by a point in space and a surface normal. In the current implementation, the normal is assumed to be in the positive y -axis direction for anastomosis, and in the negative y -axis direction for IMA takedown. In both cases the normal is perpendicular to the surgical table.