Project topics and materials
Get Complete Project Material File(s) Now! »
Robot-assisted surgery
Robot-assisted surgery, also known as robotic surgery, requires the use of a surgical robot to aid in surgical procedures. A surgical robot is a computerized system with motorized mechanism (one or more surgical arms) capable of interacting with the environment, i.e., the surgical site.
Telemanipulation
The most widely accepted commercialized robotic system is the da Vinci surgical system (Intuitive Surgical, Mountain View, CA, USA) [12]. It is a telemanipulated device placed between the surgeon and the patient which converts physical motion into electrical signals [13]. Through the comprehensive master-slave mechanism, the surgeon can operate remotely from a master console physically separated from the patient. Other surgical systems using the concept of telemanipulation include the AESOP robot (Computer Motion, Santa Barbara, CA, USA), a voice-activated robotic arm that functions as an endoscopic camera holder and the ZEUS system (Computer Motion, Santa Barbara, CA, USA), which, similarly to the daVinci system, is also a master-slave robotic surgical system. ZEUS system was discontinued in 2003, following the merger of Computer Motion with its rival Intuitive Surgical [14].
The introduction of telemanipulated surgical robots overcomes many obstacles existing in laparoscopic surgery [15–17]. Preserving most of the benefits for patients in conventional laparoscopic surgery, the improvements that telemanipulated robotic systems have brought to surgeons are obvious, namely: by translating the surgeon’s hand motions into identical instru-ment motions, the system is able to avoid the reverse-fulcrum–induced movements, to filter out hand tremors as well as to increase dexterity. The surgeon can intuitively manipulate the instruments with a proper visuo-motor coordination and improved visualization. Moreover, better postural comfort is achieved thanks to the ergonomically designed control console. Even though the use of the telemanipulation greatly enhances the surgeons’ performance, at the same time, it has raised new problems [18]. Telemanipulated electromechanical systems have large footprint, not easy to fit into the current operating theater. The installation and calibration procedures are fastidious and time-consuming. Lack of communication is another drawback since the surgeon operating on the master console is isolated from the patient. The high price tag, of course, limits the widespread use of telemanipulated systems within most healthcare organizations and points to inadequate training for certification requirements [19].
Comanipulation
Apart from the telemanipulated robots, there exists another different concept: comanipulation. It is a paradigm involving a robot and a user simultaneously manipulating a load or a tool, [20, 21]. Instead of functioning as the interface between the surgeon and the patient in teleoperation, the robot is employed as a comanipulated device, in the sense that the gesture control of the instrument is shared by the robot and the surgeon.
The currently existing commercialized comanipulated robotic systems are basically designed for specific types of surgical tasks. Acrobot (The Acrobot Company, Ltd., London, UK) is a semi-active hands-on orthopedic robotic system for total knee replacement (TKR) surgery, which allows the surgeon to safely cut the knee bones to fit a TKR prosthesis with high precision [22]. MAKOplasty (Stryker Corp, Kalamazoo, MI, USA) is a surgical robot proposed for unicompartmental knee arthroplasty (UKA) [23]. The semi-active robotic arm controlled by the surgeon can ensure implantation within acceptable limits of target specification. Robotized Surgical Assistant (ROSA, MedTech, Montpellier, France) is a robot based navigation system used for planning and implantation guidance in stereotactic placement of intracranial depth electrodes in neurosurgery [24, 25].
Apart from the commercialized systems, a number of research institutes have also exploited the idea for precise surgical tasks. Johns Hopkins University implemented the Steady Hand microsurgical robot, a specially designed cooperatively controlled robot arm for microsurgical manipulation [26]. Micron system is a hand-held actively stabilized comanipulator developed at Carnegie Mellon University [27], able to increase accuracy during eye microsurgical procedures by removing involuntary motion. Researchers in University of Paris 6 designed a compact robot named MC2E for minimally invasive surgery (MIS), which can measure the distal organ-instrument interaction with a sensor placed outside the patient [28]. A Robotic manipulator was developed in University of Leuven to assist vitreoretinal surgeons in the procedure of retinal vain cannulation [29]. This device features a new Remote Center of Motion mechanism and specially suits for highly confined working space around a fixed incision point.
Some preliminary experiments with comanipulated surgical robots showed that this con-trol form is natural and convenient for surgeons. Indeed, the comanipulation reserves the superb manual dexterity, rapid learning ability, great adaptation skills of human operators, and the same time, it combines the benefits of the robot such as the capability to enhance manipulation precision and to perform repetitive tasks without getting tired. Therefore, comanipulation can be applied to tasks that require both precise manipulation and human judgment so as to enhance gesture quality [26]. Comanipulation outperforms telemanipu-lation in the aspects of easy installation, small volume, low expense and the ability to give surgeons control sensation, even though the ergonomic comfort it offers is not comparable to the telemanipulation.
The Achilles surgical system
Nowadays, comanipulated surgical systems are principally used for high precision surgeries such as microsurgery and bone surgery. For microsurgery with a highly confined space, i.e., eye surgery, high damping is generated by comanipulators to help slow down the intended movements, maintain a stable instrument position and attenuate the surgeon’s hand tremor. For orthopedic surgery such as hip and knee surgery, predefined boundaries are usually used as geometrical guidance so that the comanipulated robots could provide motion constraint to improve surgical accuracy.
In this work, we aim at using the concept of comanipulation in the field of laparoscopic surgery. Compared with the current existing telemanipulated setups such as the da Vinci system, this system might offer the following benefits:
• High modularity. The whole system is highly modularized. Standard surgical instru-ments are able to be integrated into the system effortlessly.
• Easy installation and deinstallation. Since the robot is compact, the procedures of installation, deinstallation and integration into the application environment are easy to implement. Fastidious registration and reconfiguration of the device before or during the procedure are not required, thus shortening the operation time. In case of system failure, the surgeon can immediately change to conventional laparoscopic surgery and continue the operation.
• Better technique transferability. It does not create many troubles for surgeons to learn the robot-assisted laparoscopic surgery as long as they have mastered the conventional surgeries. Therefore, skill transferability problems are prevented [30].
• Better control sensation. Since the comanipulated system requires the surgeon stand-ing beside the patient, he or she has the feeling of being involved in the surgical control loop, which is naturally preferable. Furthermore, even though the robot can provide force on the instrument, it does not totally deprive the surgeon of the feeling of force feedback, making the operation much more reliable.
• Low expense. The cost of the comanipulated system is potentially more affordable thanks to its cheaper implementations.
A robot-comanipulated laparoscopic surgery is similar to the conventional one. During the operation, surgeons stand beside of the patient bed and maneuver the instruments. Being controlled by commands from the computer system, the robot can apply forces to the instrument so as to provide guidance to surgeon’s gestures. Fig. 1.2 is a simplified illustration of the scene of a comanipulated laparoscopic surgery. The two blue devices holding surgical instruments represent comanipulators, which we name as “Achilles”. This work mainly focuses on the research of the control of Achilles.
Questions to be solved
Lack of geometrical guidance
In a comanipulative task where the human and the robot have physical interaction, impedance control is a classical framework to use. Virtual fixtures is a common approach for motion guidance, widely used in comanipulated surgeries [2, 31–33]. The robot actively provides stiffness on the guided mechanism to limit movement into restricted region and/or guide motion along desired paths or surfaces in the workspace so that safer and faster operation can be achieved [34]. The implementation of virtual fixtures is based on a kinematic or dynamic analysis of the manipulated task, for which a precise model needs to be established. Laparoscopic surgery, unlike bone surgery, deals with soft and flexible tissues which move and deform easily during surgical procedures. Breath also leads to the motion of the organs, which forces the surgeon to aim at a dynamic target. This constant movement makes the specific geometrical and physical representation of tasks quite complex, even impossible. Therefore, although virtual fixtures is widely employed in comanipulation, for laparoscopic procedures it is not practical to apply elastic force fields (stiffness). In this work, we shift to the viscous fields, which can provide damping for the intervention of instrument dynamics [35].
Large range of motion
Literature shows that viscous force has a good performance on maintaining motions stable and precise in microsurgery [26, 27, 36]. The workspace for operating microsurgery, however, generally requires a very limited area where large motions are not allowed. Examples includes eye surgery, ear, nose and throat (ENT) surgery, dental surgery, neurosurgery, etc [37]. Slow and yet stable, accurate motions are preferred in this situation for the reason of safety. Relatively large viscous damping is thus required for higher motion resistance [38]. In laparoscopic surgery, however, the operating workspace is much larger and the range of motion varies. For precise procedures such as cutting and suturing, the principle necessity is to ensure safety and precision, thus, slow but accurate motions are necessary. On the other hand, for tasks with a relatively long distance to cover, e.g., moving the instrument tip from one position to another, high damping slows down the intended movements, leading to fatigue and more operative time consumption. Therefore, where there is no special precision requirement, large motions without external resistance are preferred. To summarize, in order to provide a level of assistance adapted to variable operating conditions (high movability or high stability), a comanipulator should be able to modify its impedance during manipulation tasks.
In this work, we propose to utilize a variable viscosity to satisfy both requirements of instrument motions, which is supposed to have the following effect: when the instrument velocity is low, the robot generates large viscous force to slow down the motion so as to stabilize the hand movement. When the instrument moves fast, no force or small force is exerted in such a way that the surgeon could manipulate the instrument easily.
Lever effect
A typical laparoscopic instrument is a long shaft with its tip inserted into the abdomen and its handle held by the surgeon outside of the patient. The trocar and the instrument form a lever, with the entry point as its fulcrum. The motion of points lying on the instrument is under the constraint of a specific relationship, which is called the lever effect (or fulcrum effect). The movement of the instrument tip and that of the handle are inversed, and their amplitude ratio for the rotational motions depends on the insertion depth of the instrument into the trocar. This might contribute to the amplification of the involuntary physical hand tremor, which would decrease surgeons’ performance, more likely to result in the increased rate of operative injuries as compared with open surgery [5]. Besides, viscosity control implemented at a certain point will lead to anisotropy for the other points on the instrument.
The position of the robot force physically imposed on the instrument, namely the con-nection point where the robot holds the instrument, is fixed. Equivalent virtual forces are generated in the instrument tip and handle, respectively, which can be computed thanks to a lever model [39]. The instrument tip directly contacts with the tissues or organs of the patient. Moreover, it influences the visual feedback information that the surgeon gains from the projected images. The handle, on the other hand, has an effect on the surgeon’s sensorimotor control system. In other words, the virtual controller implementing position may lead to different performance in terms of motion guidance. An important question is hence worthwhile to ask: which is the preferred point for the implementation of the viscosity controller to bring a desired outcome? The study of this question constitutes an important topic.
Objective of the thesis
The general objective of the work is to employ Achilles as a comanipulator to provide guidance to different surgical gestures so as to enhance the performance of surgeons during laparoscopic surgery.
Instrument gravity compensation
In a laparoscopic procedure, the surgeon does the surgery by maneuvering different types of surgical instruments rigidly inserted into the patient’s abdomen. The gravity of different instruments varies a lot. Traditional laparoscopic instruments include forceps, scissors, probes, dissectors, hooks, etc. [40]. Robotic articulated instruments are also designed and commercialized, such as JAiMY from Endocontrol and the VeSPA instruments from Intuitive Surgical Inc. Even though these motorized instruments provide more dexterity, they are much heavier (as heavy as half a kilogram) than the traditional ones due to the presence of motors and drive cables. This makes them energy-consuming to manipulate and easily leads to fatigue. More importantly, in the case of accidental instrument drop, the risk of unintentional deep insertion into the patient’s abdomen largely increases, posing grave safety hazard of tissue or organ damage during surgery. Adding gravity compensation, for both motorized instruments and for traditional ones, is hence of importance to counterbalance the instrument weight in such a way that the surgeon does not feel resistance due to instrument gravity. Therefore, in this work, the first way to achieve the objective of comanipulation gesture guidance is to compensate instrument gravity so that the surgeon would not have to carry the instrument weight.
We could add a constant value in the opposite direction of instrument gravity, as done in [41]. However, in practice, the gravity torque changes with instrument positions and orientations. Namely, when the depth of instrument insertion into the abdomen changes, its gravity generates different torques at the entry point. The angle between the instrument axis and the abdomen plane also has an influence to the gravity torque. Therefore, in this work, we use a real-time method for gravity compensation, described in detail in Chapter. 2.
Detecting trocars
The existence of trocars creates a kinematic constraint which limits the surgical instrument motion to four dofs: three independent rotations around the insertion point and one translation along the instrument longitudinal axis. When a robot is used to manipulate instruments, it is crucial to know the trocar position information with respect to the robot base body. The question of identifying this location has been the object of dense research in the past decades.
An option consists in using an external localizer. For example, in [42], a registration procedure consists, for the surgeon, in moving the endoscope around the fulcrum, while an external stereo camera pair watches the scene. The lines corresponding to the endoscope axis are extracted from several images and, thanks to a Hough transform, their intersection is computed to form a 3D trocar position estimation.
In order to avoid the use of external equipment, direct registration/installation of the robot is more often used. This is the case when using a robot exhibiting a Remote Center of Motion (RCM). Using such a mechanism requires a precise installation of the robot base body in the workspace prior to the operation, in such a way that its RCM precisely fits with patient’s entry point in order to avoid tissue damage. An example is the da Vinci robot, which is made up of four interdependent arms mounted on a single base. Each of its arms has a RCM in order to respect the constraint formed by the trocar, [30]. The robot installation procedure requires a passive arm to position the base body of each active arm. A simpler option consists in directly placing the robot on the patient, as proposed in [43] (endoscope holder) or [44] (instrument holder). Here, there is no need for an extra passive arm to position the robot base as the holder is automatically centered on the trocar. However, for all the RCM-based solutions, in the event of robot relocation during the procedure, the realignment of the robot arms to trocars requires a complete new installation process.
If the robot is to be used at several trocar locations during the same operation, a preferable approach is to use a 6-dof robot without RCM to avoid re-installation. As a price for versatility, extra work is to be done in order to guarantee that the fulcrum constraint is respected. In [45], a fully actuated 6-dof robot equipped with a force sensor is proposed. The force sensor is used both to control the movements while minimizing forces at the trocar and to estimate the fulcrum location in real time.
Exploiting a force sensor raises concerns in terms of cost, robustness, and compatibility with operating room (OR) constraints. To avoid using such an equipment, the 6-dof robot can be partially actuated and equipped with two passive joints at the wrist. The instrument can thus naturally rotate around the fulcrum point while limiting forces exerted to that point, see e.g. [46]. The AESOP robot, used in [47], makes use of such joints. To compute the online trocar point position, joint position data is collected. An algorithm that computes the best intersection between instrument axes at successive locations is used. This method does not require precise positioning of the robot, thus the setup procedure is facilitated. However, the algorithm proposed in [47] is suboptimal as it uses an average filter of a series of two-point estimates. Moreover, this solution is built on the assumption that i) the instrument is indeed inserted into a trocar; ii) the entry point does not move.
In the context of comanipulated endoscopic surgery these hypotheses do not hold: The problem is not only to localize the trocar but also to detect the trocar presence. An adapted mathematical approach, pertaining to Least Square (LS) optimization, is proposed in this work. Its practical implementation is based on rules for selecting appropriate data to feed the LS algorithm and criteria to robustly and rapidly detect the trocar presence. Chapter. 2 depicts the detailed work.
Smooth guidance of motion – variable viscosity control
Human hand tremor is an involuntary hand movement, which is approximately rhythmic and roughly sinusoidal [48]. These intrinsic limitations impede the manual positioning accuracy, even making many delicate surgical procedures impossible. The degradation of surgical performance due to noise is even greater for tasks with high level of difficulty, such as mesh alignment and suture tying [49]. The robot controller could provide smooth, tremor-free, precise positional control by sensing forces exerted by the operator on the tool. Research in the area of hand tremor suppression follows mainly two lines: teleoperated systems and cooperated system. Tremor filtering is dealt with the telerobotic technology in [50], where the motions of unstable human hands operating in the master subsystem are transfered into stable robotic arm movements. The authors in [51] designed an active hand-held microsurgical instrument named Micron for comanipulation, which implements tremor cancellation via the weighted-frequency Fourier linear combiner algorithm. In [26], an approach named steady-hand micromanipulation is developed, in which the tool is held simultaneously by the operator’s hand and a specially designed actively controlled robot arm. In this work, we expose the small motions of the surgical instrument to resistive viscous fields to filter out the hand tremor. Outside the delicate, precision-required operations such as cutting, laparoscopic procedures include also movements with large speed, i.e., transporting the instrument tip form one position to another. Damping at this stage are not desirable since it results in large force and consumes more energy whereas precision requirement is not strictly demanded. Therefore, viscosity coefficient is adjusted to be a small value in order to achieve comfortable displacement of the instrument. To conclude, the viscous controller is supposed to have the ability to adapt to different levels of motion range.
This approach of variable viscosity control was first proposed in [52] for assisting carrying objects. It is based on the experimental observation that the viscosity of a human operator’s arm drops at high velocity when collaborating with an other human operator to carry a load along a linear path. From this observation, a robot controller is designed to mimic this behavior: at low velocities, the viscosity is set high. When a velocity threshold is reached, a lower viscosity is programmed. As a result, human-robot comanipulation tasks are shown to exhibit trajectories that resemble those of human-human comanipulation tasks. However, this approach of abrupt change of viscosity results in degraded velocity control during point-to-point experiments conducted in [53]. This robot controller is then modified to be made “optimal” in [53], according to the authors. Namely, instead of abrupt changes, the viscosity coefficient follows an exponential function of time once the threshold has been reached. Resulting velocity profiles of collaborative human-robot point-to-point movements exhibit the typical bell shape of minimum jerk trajectories, which qualifies the collaborative movements as “natural”, [54]. The experimental results are appealing, however, unfortunately, this method can not be widely applied since the damping coefficient is the function of a limited time segment, not in real-time.
The variable viscosity approach was also used for more complex tasks. In a human-robot cooperative calligraphic task, [55], the viscosity coefficient is settled in proportion to the stiffness of the human operator’s arm, which is estimated in real-time. This approach requires an on-line estimator of the arm stiffness, which suffers from noise and robustness issues. In a robot assisted manual welding task, [56], the damping coefficient is a piecewise linear function of the norm of the robot velocity. The same method is employed in [57] for a task where the robot is manually guided to describe a square in the vertical plane. However, depending on the robot control parameters, this approach may lead to instability, due to the large force generated according to the linearly dropped damping coefficient.
In all these papers, the benefits of variable damping coefficient are clearly demonstrated. However, the influence of different variable viscosity coefficients on human’s hand motion profiles is not sufficiently analyzed. In the work, through theoretical analysis and practical experiments, we show that the viscosity coefficient drop creates an unstable dynamics and distorts the human natural motions. We thus introduce a second linear dynamics to slow down the viscosity coefficient variations. Namely, a first order low pass filter is used. This dynamics is experimentally shown able to stabilize the interaction and to restore the natural movements in Chapter. 3.
Lever effect
The lever effect poses the cognitive and perceptual difficulties to surgeons and is considered as the greatest ergonomic problem of the skill acquisition in laparoscopic surgery [58]. Many researchers have been focusing on the lever effect topic [4, 5, 58–61]. Some authors show interest in studying how the presence of the trocar affects the operators’ motor and cognitive sensations, such as the influence of the kinematic and dynamic transformations to the learning [60], the test of distal-shift hypotheses of extending hand movement schema to tool movement characteristics [4, 62].
Other authors focus on developing robotic devices with different kinematic designs with the aim of aiding surgeons’ manipulation through a fulcrum [39]. Some robots, such as the da Vinci surgical system, feature a remote center of motion (RCM), which serves as an invariant point fitting the fulcrum point so as to prevent damage to the patient’s tissues. A four-DOF robot exhibiting RCM is also described in [63]. This specialized kinematic design requires the robot base carefully placed in the workspace prior to instrument manipulation. A second mechanism may even be necessary for precise positioning, leading to a large system footprint. Serial robots with six-DOF are also exploited. This kind of robots possesses no “in built” invariant point and thus allows placing the robot base independently from the entry point. In the case of fully actuated robots, the kinematic constraint, i.e., the fulcrum point position, needs to be obtained in real time since this information is included in the robot’s inverse kinematics. This can be solved either through the knowledge on the fulcrum location, as proposed in [64–66], or with an additional end-effector force sensor for its estimation, as done in [67]. In the situation of partially actuated robots, which is the case of Achilles, the two passive joints at the wrist frees the orientation of the instrument around the robot end-effector point. Such mechanism allows any motion of the robot end-effector point when the instrument is inserted into the trocar and rotates around the fulcrum point, thus respecting the kinematic constraint using only four actuators.
Table of contents :
1 Introduction
1.1 Laparoscopic surgery
1.2 Robot-assisted surgery
1.2.1 Telemanipulation
1.2.2 Comanipulation
1.3 The Achilles surgical system
1.4 Questions to be solved
1.5 Objective of the thesis
1.5.1 Instrument gravity compensation
1.5.2 Detecting trocars
1.5.3 Smooth guidance of motion – variable viscosity control
1.5.4 Lever effect
1.6 Content of the work
2 A laparoscopic comanipulator
2.1 Introduction
2.2 Achilles’ characteristics
2.2.1 Mechanics
2.2.2 Kinematics
2.2.3 Actuation
2.2.4 General idea of the comanipulated system
2.3 Gravity compensation for laparoscopic instrument
2.3.1 Implementation
2.3.2 Evaluation
2.4 Robust trocar detection and localization
2.4.1 Specific aims
2.4.2 Least square algorithm
2.4.3 Trocar detection and localization
2.4.4 Experimental validation
2.5 Conclusions
3 Variable viscosity control
3.1 Introduction
3.2 Variable viscosity control
3.2.1 Basic control law
3.2.2 Theoretical analysis of instability
3.2.3 Experimental evidence of instability
3.3 Adding a dynamics to slow down viscosity variation
3.3.1 Viscosity control with filtered coefficient
3.3.2 Theoretical investigation of stability
3.3.3 Experimental evaluation of stability
3.4 Point-to-point experiment
3.4.1 Materials and methods
3.4.2 Experimental results
3.4.3 Discussion
3.5 Conclusions
4 Lever model effect
4.1 Introduction
4.2 Establishment of lever model
4.2.1 Kinematic part
4.2.2 Dynamic part
4.2.3 The computation of lever model matrix
4.2.4 Verification of lever model
4.2.5 Discussion about lever model
4.3 Viscosity control with lever model
4.4 Materials and methods
4.4.1 Experiment setup
4.4.2 Experiment protocol
4.4.3 Performance indicators
4.4.4 Data analysis
4.5 Experimental results
4.6 Discussion
4.7 Conclusions
5 Conclusions and perspectives
5.1 Conclusions
5.2 Perspectives
References
