Requirements Analysis and Preliminary Design of a Robotic Assistant for Reconstructive Microsurgery L. Vanthournhout, B. Herman, J. Duisit, F. Chˆateau, J. Szewczyk, B. Lengel´e, and B. Raucent
Abstract— Microanastomosis is a microsurgical gesture that involves suturing two very small blood vessels together. This gesture is used in many operations such as avulsed member auto-grafting, pediatric surgery, reconstructive surgery – including breast reconstruction by free flap. When vessels have diameters smaller than one millimeter, hand tremors make movements difficult to control. This paper introduces our preliminary steps towards robotic assistance for helping surgeons to perform microanastomosis in optimal conditions, in order to increase gesture quality and reliability even on smaller diameters. A general needs assessment and an experimental motion analysis were performed to define the requirements of the robot. Geometric parameters of the kinematic structure were then optimized to fulfill specific objectives. A prototype of the robot is currently being designed and built in order to providing a sufficient increase in accuracy without prolonging the duration of the procedure.
I. INTRODUCTION Each year, 500.000 women are diagnosed with breast cancer in Europe. Depending on the country, 30 to 50% of these patients are treated by mastectomy [1], [2] and a portion of them choose to have their breast reconstructed. Among breast reconstruction procedures, DIEP (Deep Inferior Epigastric Perforator) offers very good esthetic results. This technique consists of removing a skin flap with the underlying fat tissue from the patient’s abdominal wall, and grafting it into the position of the removed breast. This flap must of course be vascularized and the surgeon therefore has to suture a flap vein and artery to a breast vein and artery, respectively. This microsurgery gesture, performed under microscope, is called microanastomosis (see Fig. 1) and is also used in many other procedures in cardiac, ENT, maxillofacial, pediatric, and reconstructive surgery. This work was supported own funds from UCL and ISIR. L. Vanthournhout, B. Herman, and B. Raucent are with the Center for Reserach in Energy and Mechatronics, Institute of Mechanics, Materials, and Civil Engineering, Universit´e catholique de Louvain, Louvain-la-Neuve, Belgium, and with Louvain Bionics, Universit´e catholique de Louvain, Belgium. [email protected]. J. Duisit is with Morphology Research Group and with Experimental Surgery and Transplantation Group, Institute of Experimental and Clinical Research, Universit´e catholique de Louvain, Bruxelles, Belgium. F. Chˆateau is with Department of Plastic and Reconstructive Surgery, Cliniques universitaires Saint-Luc, Universit´e catholique de Louvain, Bruxelles, Belgium. J. Szewczyk is with Institute for Intelligent Systems and Robotics, Universit Pierre et Marie Curie Paris 06 – CNRS, Paris, France. B. Lengel´e is with Morphology Research Group, Institute of Experimental and Clinical Research, Universit´e catholique de Louvain, Bruxelles, Belgium, and with Department of Plastic and Reconstructive Surgery, Cliniques universitaires Saint-Luc, Universit´e catholique de Louvain, Bruxelles, Belgium.
978-1-4244-9270-1/15/$31.00 ©2015 IEEE
Fig. 1.
Example of microanastomosis
Currently, during a DIEP procedure, the surgeon cuts deep into the abdominal muscles and removes a piece of the rib at breast level in order to reach vessels large enough (≥1 mm in diameter) to safely make the microanastomosis. Only well-trained microsurgeons are capable of manually performing a microanastomosis at this diameter and most cannot perform this gesture on smaller vessels. Indeed, at this scale, one cannot prevent involuntary movements from happening. Fig. 2 shows the typical hand tremor when a subject tries to keep an instrument still above a target under microscope. The tremor amplitude varies with experience, fatigue, and the posture of subject, and usually has a value on the order of several hundreds of micrometers with a frequency between 6 and 15 Hz [3]–[5]. In order to improve DIEP procedure, it is necessary to increase the surgeon’s positioning accuracy. Authors estimate this accuracy at 10 µm (around 10x the best surgeon’s accuracy). An efficient way to improve accuracy is via the use
Fig. 2. (A) Error between target and ball center on the needle tip that a subject tries to keep still above a target under a microscope. (B) Typical image recorded by a microscope-mounted camcorder and processed to locate the ball center glued to the needle tip. Data is acquired at 30 Hz, spatial resolution is 1px = 6.24 µm under x4O microscope magnification.
4926
of robotic assistance with either comanipulation or teleoperation. With comanipulation, the instrument is held by both surgeon and robot. The robot filters tremors by damping the motion and the surgeon remains in direct contact with the patient. Some robots are dedicated to microsurgery such as the Micron [4], the Steady Hand [6] and the surgical manipulator developed at KUL [7]. These systems are mostly devoted to eye microsurgery during which the surgeon needs to reach a small target (e.g. a retinal blood vessel) and must keep the instrument still for several minutes. With teleoperation, the surgeon manipulates a joystick (master). The robot (slave) reproduces the joystick motion with a scaling factor to enhance accuracy and reduce the amplitude of tremors. It is usually preferred for continuous and more complex tasks. Several exploratory studies with teleoperated systems such as the da Vinci (Intuitive Surgical, Inc.) [8], [9], the RAMS workstation (JPL-Caltech) [10], the neurosurgery robotic platform (Tokyo Univ.) [11], the prototype MSR (TU Eindhoven) [12], and the RobOtol system (UPMC-CNRS Paris) [13] demonstrated the feasibility and interest of this kind of assistance for very fine gestures. However, these systems are often too cumbersome, require a long installation time and are not accurate enough to perform the complex tasks of microsurgery, mostly because of their kinematics. In addition, the teleoperation console is often placed remotely from the patient. It uses a 3D camera which does not produce an image as bright and sharp as one with a binocular microscope and it does not have a large enough magnification factor. Another limitation is that the scaling factor is often set only once during a microanastomosis procedure. However, if a large reduction increases the accuracy, it also slows down the motion and increases operation time [5], [9]–[14]. In this context, our goal is to design a teleoperated robotic assistant devoted to reconstructive microsurgery that would allow the surgeon to remain next to the patient and look at the surgical site through the existing binocular microscope so as enhance gesture accuracy when required without increasing the procedure duration. This paper reports the first steps towards this goal. First, the robot topology is introduced. Then, the motions of instruments during microsurgical gestures are analyzed to determine the workspace required for the robot. Finally, geometric parameters are optimized in accordance with different objectives in order to obtain the best solution to the surgical requirements. II. ROBOT TOPOLOGY Thanks to a users’ needs analysis performed with several surgeons teams, general requirements were set for the slave manipulator. First, the robot requires 7 degrees of freedom (DOF): Six to have an adequate dexterity and a seventh to open and close the manipulated instruments. Accuracy is mainly required for position but not for orientation, the human hand being sufficiently accurate for the latter. As a consequence, the authors have decided to use a kinematic model that fully decouples end-effector position and orientation. Three high-accuracy linear tables are placed in series
Fig. 3. wrist.
Robot topology with decoupled Cartesian carrier and spherical
along X, Y, and Z directions to translate the instrument inside the workspace. Instrument orientation is then provided by a 3-DOF spherical wrist. The axes of revolution of each wrist actuator cross at the instrument tip in order to guarantee the desired decoupling. Finally, the last rotation is aligned with the instrument axis, since the self-roll motion of the instrument is quite often used in microsurgery tasks. The robot topology is depicted in Fig. 3. This topology is based on the RobOtol system [13] dedicated to middle ear microsurgery. The robot has to be bi-manual since it is impossible to perform a microanastomosis with only one hand and also since both hands are generally used to accurate gestures. Additionally, the robot must be able to be used by both right or left-handed surgeons. Typical necessary microsurgical instruments include a straight micro-grasper, a curved micrograsper and micro-scissors. The robot has to allow for quick instruments interchangeability. III. MOTION ANALYSIS An experiment was performed to determine the linear and angular workspace required for a microsurgery procedure. A. Materials and Methods The experiment was carried out on a rat (adrenal artery: diameter ≈1.7mm) and has been approved by the Animal Ethics Review Board of the Catholic University of Louvain (2014/UCL/MD/024). Position, orientation and usage of instruments were recorded during an entire microanastomosis manual procedure, from vessel dissection to final permeability check. Three recording devices were used: • A standard camcorder captured the general scene to analyze gestures and distinguish microsurgery gestures from focus, magnification factor and instrument change gestures. • A microscope-mounted camcorder to record gestures at the microsurgery scale. R Claron Tech• A 3D tracking system (MicronTracker , nology Inc.) recorded the position and orientation of instruments. The 3D tracking system used optical markers that were mounted on the instruments. Surgeons involved in the experiment indicated that these lightweight markers did not
4927
change instruments behavior. The system accuracy was approximately 1 mm and 1.5◦ . The position and the orientation was given in a reference coordinate system defined by other optical markers which were always visible via the tracking system. The tools were calibrated before the experiment by placing them in a calibration piece which fixed them in a known position and orientation in the reference system. A manual analysis of the scene and microscope-mounted camcorders allowed for the exclusion of non-relevant data (e.g. instrument change, adjustment of the optical magnification factor on the microscope) prior to processing. Velocity and acceleration were then derived after noise filtering.
Fig. 5. Maximum and mean transverse angular velocity according to orientation.
B. Results Experimental results are depicted in Fig. 4. The position workspace of all microsurgery instruments used can be entirely included in a cuboid of 40×50×40mm3 when the right side is brought back to the left side by mirror symmetry. A maximum velocity of 52mm/s includes 99% of relevant data and 100mm/s includes 99.9%. Likewise, a maximum acceleration of 530mm/s2 includes 99% of relevant data and 1030mm/s2 includes 99.9%. As regards orientation, considering the robot topology, the instrument self-rotation (longitudinal) axis is decoupled from the two other angular DOFs. Angular range of the selfaxis exceeds 360◦ during microsurgery since the surgeon very often rotates the instrument. Around this self-axis, a maximum angular velocity of 74◦ /s includes 99% of relevant data and 700◦ /s includes 99.9%. Likewise, a maximum angular acceleration of 520◦ /s2 includes 99% of relevant data and 10700◦ /s2 includes 99.9%. The remaining two orientations – called transverse hereafter – can be depicted by points on the surface of a sphere centered on the instrument tip, each point representing the intersection of the instrument self-axis with the sphere. The right side is also brought back to the left side by mirror symmetry. All orientations can be included in an envelope as shown in Fig. 4B. A maximum angular velocity of 28◦ /s
Fig. 4.
includes 99% of relevant data and 56◦ /s includes 99.9%. Likewise, a maximum angular acceleration of 240◦ /s2 includes 99% of relevant data and 480◦ /s2 includes 99.9%. Transverse angular velocity was further analyzed. Figure 5 shows that the maximum velocity is reached in most of the orientation workspace. Conversely, mean angular velocity seems to be slower at the workspace center whereas gestures performed near the workspace border are often faster. C. Discussion Although this experiment was performed by two microsurgeons, data from the 3D tracking system were analyzed in detail for one surgeon only. Scene and microscope recordings from the second surgeon were analyzed to confirm that no other movement was performed. Yet, if these results give a fair idea of the gestures performed during microanastomosis, they still might not be representative of all microsurgeons, and additional trials could be performed if required. Moreover, with robotic assistance, gestures could be modified. The results reported above allow to be set the requirements for the three linear tables of the Cartesian carrier. The workspace must be larger than 40×50×40mm3 and each axis must be able to reach a velocity of 100mm/s and an acceleration of 1030mm/s2 . Required angular velocity, acceleration,
(A) Experimental environment. (B) Instruments orientation. (C) Instruments tip position.
4928
and torque for each revolute joint, however, depend on the spherical wrist weight and geometric parameters, the latter determined by an optimization process described below. Finally, transverse angular velocity has no favourite direction, and a sufficient dexterity is therefore required over the whole workspace. IV. GEOMETRIC OPTIMIZATION Now that the required workspace is known from the previous section, spherical wrist geometric parameters can be optimized as follows. A. Materials and Methods The spherical wrist dimensions can be defined by 7 parameters that were optimized (see Fig. 6A): • r1 , α1 and β1 define the position and orientation of first actuated joint with respect to a reference frame attached to the Cartesian carrier and centered at the remote renter-of-motion (RCM), expressed in spherical coordinates (resp. radius, longitude and latitude); • α2 defines the angle between the first and the second joints, and L2 defines the distance between the RCM and the second joint; • α3 defines the angle between the second and the third joints, and L3 defines the distance between the RCM and the third joint. An optimization algorithm summarized in Fig. 7 explores the solutions space systematically. If each pi Q7 parameter i covers [xi ; yi ] by step of si , there are i=1 ( xis−y + 1) i solutions to test. Optimization was performed with Matlab 2013a (MathWorks) on 3 computers (Intel Core 2 Q9500 @2.38 GHz, RAM 4 GB) in parallel, under Windows 7 64 bits. Each optimization run lasted for several days (around a hundred hours). The robot must be able to cover the whole workspace without colliding with the environment (including the robot itself and a second robot for bi-manual gestures) and without passing under the microscope field of view. Kinematic singularities have to be located as far as possible from the workspace and the distance from environment must be as large as possible. Finally, robot motion should not disturb
Fig. 7.
Optimization scheme
a second pair of slave arms that could be used by a second surgeon in front of the main surgeon. These requirements are classified in constraints and objectives. For each tested solution, the algorithm checks whether it fulfills all constraints. The solution is either rejected if at least one constraint is not satisfied, or ranked according to its capacities to meet the objectives (see Fig. 7). B. Results After a first optimization run, results showed that no solution could leave enough room for a second bi-manual device. Several options were then considered to solve this problem. One possibility was to reduce the workspace. This could lead to solutions with other acceptable characteristics but may also cause the surgeon to feel restricted in his gestures. Another possibility was to add some redundancy in the wrist topology so as to cover the whole workspace with a smaller footprint. However, this solution increases design complexity and cost. A last possibility was to eliminate the
Fig. 6. (A) Geometric parameters to optimize. (Surgeon is on the positive y axis.) (B) Parameters range and step of the first and the last optimisation. (C) Pareto front of optimization results. (D) Simplified CAD representation of the selected solution.
4929
second bi-manual system. In the operating room, surgeons usually work in tandem and this choice will be restrictive. Nevertheless, the presence of an assistant surgeon is not obligatory and his gestures usually require less fine dexterity and can thus instead be performed by hand. In addition, the first system prototype will mostly be used for experimental performance assessment in laboratory environment without a second surgeon. Since workspace use during robotic teleoperation might differ from the preliminary assessment with direct manual gestures, the final requirements for a device closer to clinical use might change and, at this juncture, the optimization can then be refined. For all these reasons, it was decided to remove the second bi-manual system from objectives. Optimization was then performed in several runs – first, with a large range for each parameter and rough steps; then, with finer steps in a smaller interval of interest (see table of Fig. 6B). The final optimization results are depicted in Fig. 6C. Each point on this curve (Pareto front) corresponds to a set of geometric parameters, and the selected solution is higlighted. C. Discussion The solution choice in Pareto front of better solution can be seen as subjective. As joints are modeled as pure axes in the optimization process, the distance to the environment must be larger than 20mm for placing actuators or means of transmission. Authors selected a solution that offers a good balance between dexterity and distance to the environment. The final robot parameters are r1 =190mm, α1 =-60◦ , β1 =58◦ , L2 =210mm, α2 =46◦ , L3 =160mm and α3 =45◦ . A schematic CAD representation of this solution is shown in Fig. 6D. V. CONCLUSIONS AND FUTURE WORK With a suitable robotic assistance, surgeons could perform microanastomoses more reliably and on blood vessels with a smaller diameter, following the global trend to decrease invasiveness of resconstructive miscrosurgery procedures. In this paper, authors set requirements for such a system based on the analysis of existing solutions, on meetings with several surgical teams, and on experimental assessment of manual gestures workspace. This requirements analysis allowed to select a suitable topology composed of a highaccuracy Cartesian carrier and a decoupled spherical wrist allowing large angular motions of the instrument. Geometric parameters of this wrist were then optimized according to its dexterity and its environment. A prototype is currently being designed and built. Special attention will then be paid to the control interface, since interaction modes between surgeon and robot are a decisive factor in the system performance and acceptance. In particular, several adaptive strategies will be proposed for providing at each moment the adequate scaling factor between master joystick and slave effector. The impact of these strategies on system performance will be studied regarding criteria such as ergonomics, usability, gestures accuracy, and task duration.
ACKNOWLEDGMENT Authors are grateful to the members of the reconstructive, maxillofacial, and otolaryngology microsurgery teams of the Cliniques universitaires Saint-Luc who took part in discussions, observations and experiments. Authors would also like to thank Prof. Pierre Gianello and his team from Experimental Surgery Laboratory (CHEX, UCL), Prof. JeanPaul Dehoux, and all colleagues from the CEREM and ISIR for their help and support in this project. R EFERENCES [1] “Globocan: Breast cancer, estimated incidence, mortality and prevalence worldwide in 2012,” Internet:http://globocan.iarc.fr/Pages/fact sheets cancer.aspx, [2015-03-27]. [2] A. Richard, “Chirurgie conservatrice du cancer du sein : de grandes disparit´es selon les pays,” Internet: http: //sante-guerir.notrefamille.com/sante-a-z/chirurgie-conservatrice-du-/ cancer-du-sein-de-grandes-disparites-selon-les-pays-o58915.html, [2015-03-27]. [3] K. Veluvolu and W. Ang, “Estimation and filtering of physiological tremor for real-time compensation in surgical robotics applications,” The International Journal of Medical Robotics and Computer Assisted Surgery, vol. 6, no. 3, pp. 334–342, 2010. [4] R. MacLachlan, B. Becker, J. Cuevas Tabare?s, G. Podnar, L. Lobes, and C. Riviere, “Micron: An actively stabilized handheld tool for microsurgery,” Robotics, IEEE Transactions on, vol. 28, no. 1, pp. 195–212, Feb 2012. [5] B. Safwat, E. Su, R. Gassert, C. Teo, and E. Burdet, “The role of posture, magnification, and grip force on microscopic accuracy,” Ann Biomed Eng, vol. 37, no. 5, pp. 997–1006, Mar. 2009. [6] B. Mitchell, J. Koo, M. Iordachita, P. Kazanzides, A. Kapoor, J. Handa, G. Hager, and R. Taylor, “Development and application of a new steady-hand manipulator for retinal surgery,” in Robotics and Automation, 2007 IEEE International Conference on. IEEE, 2007, pp. 623– 629. [7] A. Gijbels, N. Wouters, P. Stalmans, H. Van Brussel, D. Reynaerts, and E. Poorten, “Design and realisation of a novel robotic manipulator for retinal surgery,” in Intelligent Robots and Systems (IROS), 2013 IEEE/RSJ International Conference on. IEEE, 2013, pp. 3598–3603. [8] R. Katz, G. Rosson, J. Taylor, and N. Singh, “Robotics in microsurgery: use of a surgical robot to perform a free flap in a pig,” Microsurgery, vol. 25, no. 7, pp. 566–569, Sep. 2005. [9] P. Liverneaux, S. Berner, M. Bednar, S. Parekattil, G. Ruggiero, and J. Selber, Telemicrosurgery: robot assisted microsurgery. Springer, 2013. [10] P. Le Roux, H. Das, S. Esquenazi, and J. Kelly, “Robot-assisted microsurgery: a feasibility study in the rat,” Neurosurgery, vol. 48, no. 3, pp. 584–589, Jun. 2001. [11] M. Mitsuishi, A. Morita, N. Sugita, S. Sora, R. Mochizuki, K. Tanimoto, Y. Baek, H. Takahashi, and K. Harada, “Master-slave robotic platform and its feasibility study for micro-neurosurgery,” Int J Med Robot Comput Assist Surg, vol. 9, no. 2, pp. 180–189, Jun. 2013. [12] R. Cau, “Design and realization of a master-slave system for reconstructive microsurgery,” Ph.D. dissertation, Eindhoven University of Technology, Feb. 2014. [13] M. Miroir, Y. Nguyen, J. Szewczyk, O. Sterkers, and A. Grayeli, “Design, kinematic optimization, and evaluation of a teleoperated system for middle ear microsurgery,” The Scientific World Journal, vol. 2012, Jun. 2012. [14] S. Salcudean, S. Ku, and G. Bell, “Performance measurement in scaled teleoperation for microsurgery,” in CVRMed-MRCAS’97, Grenoble, Mar. 1997, pp. 789–798.
4930
Abstract— Microanastomosis is a microsurgical gesture that involves suturing two very small blood vessels together. This gesture is used in many operations such as avulsed member auto-grafting, pediatric surgery, reconstructive surgery – including breast reconstruction by free flap. When vessels have diameters smaller than one millimeter, hand tremors make movements difficult to control. This paper introduces our preliminary steps towards robotic assistance for helping surgeons to perform microanastomosis in optimal conditions, in order to increase gesture quality and reliability even on smaller diameters. A general needs assessment and an experimental motion analysis were performed to define the requirements of the robot. Geometric parameters of the kinematic structure were then optimized to fulfill specific objectives. A prototype of the robot is currently being designed and built in order to providing a sufficient increase in accuracy without prolonging the duration of the procedure.
I. INTRODUCTION Each year, 500.000 women are diagnosed with breast cancer in Europe. Depending on the country, 30 to 50% of these patients are treated by mastectomy [1], [2] and a portion of them choose to have their breast reconstructed. Among breast reconstruction procedures, DIEP (Deep Inferior Epigastric Perforator) offers very good esthetic results. This technique consists of removing a skin flap with the underlying fat tissue from the patient’s abdominal wall, and grafting it into the position of the removed breast. This flap must of course be vascularized and the surgeon therefore has to suture a flap vein and artery to a breast vein and artery, respectively. This microsurgery gesture, performed under microscope, is called microanastomosis (see Fig. 1) and is also used in many other procedures in cardiac, ENT, maxillofacial, pediatric, and reconstructive surgery. This work was supported own funds from UCL and ISIR. L. Vanthournhout, B. Herman, and B. Raucent are with the Center for Reserach in Energy and Mechatronics, Institute of Mechanics, Materials, and Civil Engineering, Universit´e catholique de Louvain, Louvain-la-Neuve, Belgium, and with Louvain Bionics, Universit´e catholique de Louvain, Belgium. [email protected]. J. Duisit is with Morphology Research Group and with Experimental Surgery and Transplantation Group, Institute of Experimental and Clinical Research, Universit´e catholique de Louvain, Bruxelles, Belgium. F. Chˆateau is with Department of Plastic and Reconstructive Surgery, Cliniques universitaires Saint-Luc, Universit´e catholique de Louvain, Bruxelles, Belgium. J. Szewczyk is with Institute for Intelligent Systems and Robotics, Universit Pierre et Marie Curie Paris 06 – CNRS, Paris, France. B. Lengel´e is with Morphology Research Group, Institute of Experimental and Clinical Research, Universit´e catholique de Louvain, Bruxelles, Belgium, and with Department of Plastic and Reconstructive Surgery, Cliniques universitaires Saint-Luc, Universit´e catholique de Louvain, Bruxelles, Belgium.
978-1-4244-9270-1/15/$31.00 ©2015 IEEE
Fig. 1.
Example of microanastomosis
Currently, during a DIEP procedure, the surgeon cuts deep into the abdominal muscles and removes a piece of the rib at breast level in order to reach vessels large enough (≥1 mm in diameter) to safely make the microanastomosis. Only well-trained microsurgeons are capable of manually performing a microanastomosis at this diameter and most cannot perform this gesture on smaller vessels. Indeed, at this scale, one cannot prevent involuntary movements from happening. Fig. 2 shows the typical hand tremor when a subject tries to keep an instrument still above a target under microscope. The tremor amplitude varies with experience, fatigue, and the posture of subject, and usually has a value on the order of several hundreds of micrometers with a frequency between 6 and 15 Hz [3]–[5]. In order to improve DIEP procedure, it is necessary to increase the surgeon’s positioning accuracy. Authors estimate this accuracy at 10 µm (around 10x the best surgeon’s accuracy). An efficient way to improve accuracy is via the use
Fig. 2. (A) Error between target and ball center on the needle tip that a subject tries to keep still above a target under a microscope. (B) Typical image recorded by a microscope-mounted camcorder and processed to locate the ball center glued to the needle tip. Data is acquired at 30 Hz, spatial resolution is 1px = 6.24 µm under x4O microscope magnification.
4926
of robotic assistance with either comanipulation or teleoperation. With comanipulation, the instrument is held by both surgeon and robot. The robot filters tremors by damping the motion and the surgeon remains in direct contact with the patient. Some robots are dedicated to microsurgery such as the Micron [4], the Steady Hand [6] and the surgical manipulator developed at KUL [7]. These systems are mostly devoted to eye microsurgery during which the surgeon needs to reach a small target (e.g. a retinal blood vessel) and must keep the instrument still for several minutes. With teleoperation, the surgeon manipulates a joystick (master). The robot (slave) reproduces the joystick motion with a scaling factor to enhance accuracy and reduce the amplitude of tremors. It is usually preferred for continuous and more complex tasks. Several exploratory studies with teleoperated systems such as the da Vinci (Intuitive Surgical, Inc.) [8], [9], the RAMS workstation (JPL-Caltech) [10], the neurosurgery robotic platform (Tokyo Univ.) [11], the prototype MSR (TU Eindhoven) [12], and the RobOtol system (UPMC-CNRS Paris) [13] demonstrated the feasibility and interest of this kind of assistance for very fine gestures. However, these systems are often too cumbersome, require a long installation time and are not accurate enough to perform the complex tasks of microsurgery, mostly because of their kinematics. In addition, the teleoperation console is often placed remotely from the patient. It uses a 3D camera which does not produce an image as bright and sharp as one with a binocular microscope and it does not have a large enough magnification factor. Another limitation is that the scaling factor is often set only once during a microanastomosis procedure. However, if a large reduction increases the accuracy, it also slows down the motion and increases operation time [5], [9]–[14]. In this context, our goal is to design a teleoperated robotic assistant devoted to reconstructive microsurgery that would allow the surgeon to remain next to the patient and look at the surgical site through the existing binocular microscope so as enhance gesture accuracy when required without increasing the procedure duration. This paper reports the first steps towards this goal. First, the robot topology is introduced. Then, the motions of instruments during microsurgical gestures are analyzed to determine the workspace required for the robot. Finally, geometric parameters are optimized in accordance with different objectives in order to obtain the best solution to the surgical requirements. II. ROBOT TOPOLOGY Thanks to a users’ needs analysis performed with several surgeons teams, general requirements were set for the slave manipulator. First, the robot requires 7 degrees of freedom (DOF): Six to have an adequate dexterity and a seventh to open and close the manipulated instruments. Accuracy is mainly required for position but not for orientation, the human hand being sufficiently accurate for the latter. As a consequence, the authors have decided to use a kinematic model that fully decouples end-effector position and orientation. Three high-accuracy linear tables are placed in series
Fig. 3. wrist.
Robot topology with decoupled Cartesian carrier and spherical
along X, Y, and Z directions to translate the instrument inside the workspace. Instrument orientation is then provided by a 3-DOF spherical wrist. The axes of revolution of each wrist actuator cross at the instrument tip in order to guarantee the desired decoupling. Finally, the last rotation is aligned with the instrument axis, since the self-roll motion of the instrument is quite often used in microsurgery tasks. The robot topology is depicted in Fig. 3. This topology is based on the RobOtol system [13] dedicated to middle ear microsurgery. The robot has to be bi-manual since it is impossible to perform a microanastomosis with only one hand and also since both hands are generally used to accurate gestures. Additionally, the robot must be able to be used by both right or left-handed surgeons. Typical necessary microsurgical instruments include a straight micro-grasper, a curved micrograsper and micro-scissors. The robot has to allow for quick instruments interchangeability. III. MOTION ANALYSIS An experiment was performed to determine the linear and angular workspace required for a microsurgery procedure. A. Materials and Methods The experiment was carried out on a rat (adrenal artery: diameter ≈1.7mm) and has been approved by the Animal Ethics Review Board of the Catholic University of Louvain (2014/UCL/MD/024). Position, orientation and usage of instruments were recorded during an entire microanastomosis manual procedure, from vessel dissection to final permeability check. Three recording devices were used: • A standard camcorder captured the general scene to analyze gestures and distinguish microsurgery gestures from focus, magnification factor and instrument change gestures. • A microscope-mounted camcorder to record gestures at the microsurgery scale. R Claron Tech• A 3D tracking system (MicronTracker , nology Inc.) recorded the position and orientation of instruments. The 3D tracking system used optical markers that were mounted on the instruments. Surgeons involved in the experiment indicated that these lightweight markers did not
4927
change instruments behavior. The system accuracy was approximately 1 mm and 1.5◦ . The position and the orientation was given in a reference coordinate system defined by other optical markers which were always visible via the tracking system. The tools were calibrated before the experiment by placing them in a calibration piece which fixed them in a known position and orientation in the reference system. A manual analysis of the scene and microscope-mounted camcorders allowed for the exclusion of non-relevant data (e.g. instrument change, adjustment of the optical magnification factor on the microscope) prior to processing. Velocity and acceleration were then derived after noise filtering.
Fig. 5. Maximum and mean transverse angular velocity according to orientation.
B. Results Experimental results are depicted in Fig. 4. The position workspace of all microsurgery instruments used can be entirely included in a cuboid of 40×50×40mm3 when the right side is brought back to the left side by mirror symmetry. A maximum velocity of 52mm/s includes 99% of relevant data and 100mm/s includes 99.9%. Likewise, a maximum acceleration of 530mm/s2 includes 99% of relevant data and 1030mm/s2 includes 99.9%. As regards orientation, considering the robot topology, the instrument self-rotation (longitudinal) axis is decoupled from the two other angular DOFs. Angular range of the selfaxis exceeds 360◦ during microsurgery since the surgeon very often rotates the instrument. Around this self-axis, a maximum angular velocity of 74◦ /s includes 99% of relevant data and 700◦ /s includes 99.9%. Likewise, a maximum angular acceleration of 520◦ /s2 includes 99% of relevant data and 10700◦ /s2 includes 99.9%. The remaining two orientations – called transverse hereafter – can be depicted by points on the surface of a sphere centered on the instrument tip, each point representing the intersection of the instrument self-axis with the sphere. The right side is also brought back to the left side by mirror symmetry. All orientations can be included in an envelope as shown in Fig. 4B. A maximum angular velocity of 28◦ /s
Fig. 4.
includes 99% of relevant data and 56◦ /s includes 99.9%. Likewise, a maximum angular acceleration of 240◦ /s2 includes 99% of relevant data and 480◦ /s2 includes 99.9%. Transverse angular velocity was further analyzed. Figure 5 shows that the maximum velocity is reached in most of the orientation workspace. Conversely, mean angular velocity seems to be slower at the workspace center whereas gestures performed near the workspace border are often faster. C. Discussion Although this experiment was performed by two microsurgeons, data from the 3D tracking system were analyzed in detail for one surgeon only. Scene and microscope recordings from the second surgeon were analyzed to confirm that no other movement was performed. Yet, if these results give a fair idea of the gestures performed during microanastomosis, they still might not be representative of all microsurgeons, and additional trials could be performed if required. Moreover, with robotic assistance, gestures could be modified. The results reported above allow to be set the requirements for the three linear tables of the Cartesian carrier. The workspace must be larger than 40×50×40mm3 and each axis must be able to reach a velocity of 100mm/s and an acceleration of 1030mm/s2 . Required angular velocity, acceleration,
(A) Experimental environment. (B) Instruments orientation. (C) Instruments tip position.
4928
and torque for each revolute joint, however, depend on the spherical wrist weight and geometric parameters, the latter determined by an optimization process described below. Finally, transverse angular velocity has no favourite direction, and a sufficient dexterity is therefore required over the whole workspace. IV. GEOMETRIC OPTIMIZATION Now that the required workspace is known from the previous section, spherical wrist geometric parameters can be optimized as follows. A. Materials and Methods The spherical wrist dimensions can be defined by 7 parameters that were optimized (see Fig. 6A): • r1 , α1 and β1 define the position and orientation of first actuated joint with respect to a reference frame attached to the Cartesian carrier and centered at the remote renter-of-motion (RCM), expressed in spherical coordinates (resp. radius, longitude and latitude); • α2 defines the angle between the first and the second joints, and L2 defines the distance between the RCM and the second joint; • α3 defines the angle between the second and the third joints, and L3 defines the distance between the RCM and the third joint. An optimization algorithm summarized in Fig. 7 explores the solutions space systematically. If each pi Q7 parameter i covers [xi ; yi ] by step of si , there are i=1 ( xis−y + 1) i solutions to test. Optimization was performed with Matlab 2013a (MathWorks) on 3 computers (Intel Core 2 Q9500 @2.38 GHz, RAM 4 GB) in parallel, under Windows 7 64 bits. Each optimization run lasted for several days (around a hundred hours). The robot must be able to cover the whole workspace without colliding with the environment (including the robot itself and a second robot for bi-manual gestures) and without passing under the microscope field of view. Kinematic singularities have to be located as far as possible from the workspace and the distance from environment must be as large as possible. Finally, robot motion should not disturb
Fig. 7.
Optimization scheme
a second pair of slave arms that could be used by a second surgeon in front of the main surgeon. These requirements are classified in constraints and objectives. For each tested solution, the algorithm checks whether it fulfills all constraints. The solution is either rejected if at least one constraint is not satisfied, or ranked according to its capacities to meet the objectives (see Fig. 7). B. Results After a first optimization run, results showed that no solution could leave enough room for a second bi-manual device. Several options were then considered to solve this problem. One possibility was to reduce the workspace. This could lead to solutions with other acceptable characteristics but may also cause the surgeon to feel restricted in his gestures. Another possibility was to add some redundancy in the wrist topology so as to cover the whole workspace with a smaller footprint. However, this solution increases design complexity and cost. A last possibility was to eliminate the
Fig. 6. (A) Geometric parameters to optimize. (Surgeon is on the positive y axis.) (B) Parameters range and step of the first and the last optimisation. (C) Pareto front of optimization results. (D) Simplified CAD representation of the selected solution.
4929
second bi-manual system. In the operating room, surgeons usually work in tandem and this choice will be restrictive. Nevertheless, the presence of an assistant surgeon is not obligatory and his gestures usually require less fine dexterity and can thus instead be performed by hand. In addition, the first system prototype will mostly be used for experimental performance assessment in laboratory environment without a second surgeon. Since workspace use during robotic teleoperation might differ from the preliminary assessment with direct manual gestures, the final requirements for a device closer to clinical use might change and, at this juncture, the optimization can then be refined. For all these reasons, it was decided to remove the second bi-manual system from objectives. Optimization was then performed in several runs – first, with a large range for each parameter and rough steps; then, with finer steps in a smaller interval of interest (see table of Fig. 6B). The final optimization results are depicted in Fig. 6C. Each point on this curve (Pareto front) corresponds to a set of geometric parameters, and the selected solution is higlighted. C. Discussion The solution choice in Pareto front of better solution can be seen as subjective. As joints are modeled as pure axes in the optimization process, the distance to the environment must be larger than 20mm for placing actuators or means of transmission. Authors selected a solution that offers a good balance between dexterity and distance to the environment. The final robot parameters are r1 =190mm, α1 =-60◦ , β1 =58◦ , L2 =210mm, α2 =46◦ , L3 =160mm and α3 =45◦ . A schematic CAD representation of this solution is shown in Fig. 6D. V. CONCLUSIONS AND FUTURE WORK With a suitable robotic assistance, surgeons could perform microanastomoses more reliably and on blood vessels with a smaller diameter, following the global trend to decrease invasiveness of resconstructive miscrosurgery procedures. In this paper, authors set requirements for such a system based on the analysis of existing solutions, on meetings with several surgical teams, and on experimental assessment of manual gestures workspace. This requirements analysis allowed to select a suitable topology composed of a highaccuracy Cartesian carrier and a decoupled spherical wrist allowing large angular motions of the instrument. Geometric parameters of this wrist were then optimized according to its dexterity and its environment. A prototype is currently being designed and built. Special attention will then be paid to the control interface, since interaction modes between surgeon and robot are a decisive factor in the system performance and acceptance. In particular, several adaptive strategies will be proposed for providing at each moment the adequate scaling factor between master joystick and slave effector. The impact of these strategies on system performance will be studied regarding criteria such as ergonomics, usability, gestures accuracy, and task duration.
ACKNOWLEDGMENT Authors are grateful to the members of the reconstructive, maxillofacial, and otolaryngology microsurgery teams of the Cliniques universitaires Saint-Luc who took part in discussions, observations and experiments. Authors would also like to thank Prof. Pierre Gianello and his team from Experimental Surgery Laboratory (CHEX, UCL), Prof. JeanPaul Dehoux, and all colleagues from the CEREM and ISIR for their help and support in this project. R EFERENCES [1] “Globocan: Breast cancer, estimated incidence, mortality and prevalence worldwide in 2012,” Internet:http://globocan.iarc.fr/Pages/fact sheets cancer.aspx, [2015-03-27]. [2] A. Richard, “Chirurgie conservatrice du cancer du sein : de grandes disparit´es selon les pays,” Internet: http: //sante-guerir.notrefamille.com/sante-a-z/chirurgie-conservatrice-du-/ cancer-du-sein-de-grandes-disparites-selon-les-pays-o58915.html, [2015-03-27]. [3] K. Veluvolu and W. Ang, “Estimation and filtering of physiological tremor for real-time compensation in surgical robotics applications,” The International Journal of Medical Robotics and Computer Assisted Surgery, vol. 6, no. 3, pp. 334–342, 2010. [4] R. MacLachlan, B. Becker, J. Cuevas Tabare?s, G. Podnar, L. Lobes, and C. Riviere, “Micron: An actively stabilized handheld tool for microsurgery,” Robotics, IEEE Transactions on, vol. 28, no. 1, pp. 195–212, Feb 2012. [5] B. Safwat, E. Su, R. Gassert, C. Teo, and E. Burdet, “The role of posture, magnification, and grip force on microscopic accuracy,” Ann Biomed Eng, vol. 37, no. 5, pp. 997–1006, Mar. 2009. [6] B. Mitchell, J. Koo, M. Iordachita, P. Kazanzides, A. Kapoor, J. Handa, G. Hager, and R. Taylor, “Development and application of a new steady-hand manipulator for retinal surgery,” in Robotics and Automation, 2007 IEEE International Conference on. IEEE, 2007, pp. 623– 629. [7] A. Gijbels, N. Wouters, P. Stalmans, H. Van Brussel, D. Reynaerts, and E. Poorten, “Design and realisation of a novel robotic manipulator for retinal surgery,” in Intelligent Robots and Systems (IROS), 2013 IEEE/RSJ International Conference on. IEEE, 2013, pp. 3598–3603. [8] R. Katz, G. Rosson, J. Taylor, and N. Singh, “Robotics in microsurgery: use of a surgical robot to perform a free flap in a pig,” Microsurgery, vol. 25, no. 7, pp. 566–569, Sep. 2005. [9] P. Liverneaux, S. Berner, M. Bednar, S. Parekattil, G. Ruggiero, and J. Selber, Telemicrosurgery: robot assisted microsurgery. Springer, 2013. [10] P. Le Roux, H. Das, S. Esquenazi, and J. Kelly, “Robot-assisted microsurgery: a feasibility study in the rat,” Neurosurgery, vol. 48, no. 3, pp. 584–589, Jun. 2001. [11] M. Mitsuishi, A. Morita, N. Sugita, S. Sora, R. Mochizuki, K. Tanimoto, Y. Baek, H. Takahashi, and K. Harada, “Master-slave robotic platform and its feasibility study for micro-neurosurgery,” Int J Med Robot Comput Assist Surg, vol. 9, no. 2, pp. 180–189, Jun. 2013. [12] R. Cau, “Design and realization of a master-slave system for reconstructive microsurgery,” Ph.D. dissertation, Eindhoven University of Technology, Feb. 2014. [13] M. Miroir, Y. Nguyen, J. Szewczyk, O. Sterkers, and A. Grayeli, “Design, kinematic optimization, and evaluation of a teleoperated system for middle ear microsurgery,” The Scientific World Journal, vol. 2012, Jun. 2012. [14] S. Salcudean, S. Ku, and G. Bell, “Performance measurement in scaled teleoperation for microsurgery,” in CVRMed-MRCAS’97, Grenoble, Mar. 1997, pp. 789–798.
4930
