References

RESEARCH STRATEGY

SIGNIFICANCE: There is broad consensus that (a) the future of minimally invasive surgery lies in natural orifice/single port access [49] and that (b) robotics is the key technology to bring us to this future. Our group has pioneered single-port robotic subxiphoid interventions on the beating heart since 2001[73][62] [77][35][37][42][44][23]. Other telerobotic mechanisms have been developed for reaching inside confined surgical spaces though a single port: snake robots [4][6][7][8][9][11][15], wireless surgical capsules[22], inchworm robots (such as our own Heartlander [23]), and flexible needles [24][25][27]. Several of these mechanisms are only capable of monitoring confined surgical spaces [22], or injecting drugs [24][25][27]. We advocate the use of snake robots because can reach inside confined surgical spaces through a small single port and can perform surgical tasks.

As Taylor described [51], a robotic instrument for minimally invasive surgery (MIS) must have sufficient dexterity, strength, and accuracy for its intended use. Both complex tissue manipulation (e.g. deforming, dissecting, suturing, etc.) and navigation require tip dexterity, especially in confined surgical spaces. Tissue manipulation requires the “right” amount of strength, i.e., stiffness, at the tip: low tip stiffness may not be sufficient to manipulate tissue, whereas high tip stiffness runs the risk of damaging delicate tissues while navigating to a target site. Therefore it is critical to modulate the stiffness of the tip of a surgical snake robot according to the surgical task and the location of the tip. Unfortunately, either snake robots have high stiffness and low dexterity, such as the serpentine robots consisting of a series of rigid links[43], or they have high dexterity and low stiffness [46][47]., such as the continuum robots [3] whose flexible tubular structures assume the shape of smooth curve whose entire curvature can be controlled (Error: Referencesource not found). Continuum robots include steerable catheters [4][10], super elastic backbone robots[6] and concentric tube steerable cannula [8][11].

Our group has a vast experience in building snake robots on both sides of the spectrum. The highly articulated robotic probe (HARP) developed by our group for MIS (Error:Reference source not found-left)[73] is a serpentine robot with 30 links and the outside diameter of 11mm. The HARP can have any desired curve with minimum radius of curvature of 35mm. A key feature of the HARP is that it uses conventional actuation, unlike other research platforms based on shape memory alloy actuators [52][31]., which suffer from a slow response time, low contact force, and excessive heat The HARP has been demonstrated in 30 pigs, 2 cadavers and 3 live humans [CITE]. Since the HARP is made of rigid links, it has high stiffness but low dexterity. On the other hand, we have developed steerable cannulas that provide high tip dexterity inside confined spaces, but alas, by design are flexible and therefore have limited strength. We hypothesize that, in order to enable complex single-port procedures in hard-to-reach locations inside the body (e.g. epicardial interventions on the beating heart), the introduction of snake robots with high dexterity and sufficient stiffness may provide an elegant solution. Hence, our proposal addresses a critical barrier to progress in the field of robotic surgery.

Constructing the mechanism, however, is not enough. The physician requires a sense of awareness of the operating site in order to both access it and deliver therapies. Direct video assistance has limited role for navigation in virtual spaces (e.g. pericardial sac) due to the confined physical space. Conventional image-guidance techniques for MIS provide minimal information to the user: they are most often limited to either 1) 2D projective views of the operation, such as fluoroscopy [CITE], 2) a simplified rendered visualization that shows a tracking device registered to preoperative images [CITE], or 3) a fusion of ultrasound with 3D reconstructed images [CITE]. All of these methods are limited in the amount of information that can be conveyed to the surgeon. These existing methods can also be inaccurate and unintuitive to use during a surgical operation. We believe that MIS would benefit from a new image-guidance method that fuses multiple sensing modalities, that estimates the inherent motion of internal organs, and that incorporates constraints imposed on the robot’s path in a stochastic filtering framework, all while simplifying the graphical interface. Showing the precise location and shape of the surgical robot in relation to a 4D 3D rendered anatomical target would be invaluable for planning and feedback. The 34D modeklmodel, specifically, would provide a more accurate view of the operating site. Therefore, image-guided snake robots will provide a truly elegant solution.

Figure 1: (left) HARP (right) Steerable cannula

As an exemplary application of single port intervention that requires a snake robot with high dexterity and sufficient stiffness, we propose to focus the current research proposal on radiofrequency (RF) ablation for atrial fibrillation (AF). AF is the most common cardiac arrhythmia affecting >4 million Americans, leading to death, stroke, heart failure and reduced quality of life. A novel minimally invasive epicardial ablation approach has been introduced and results in 94% freedom from recurrent AF without anti arrhythmic drugs[1]. However, this procedure has limited application because it requires bilateral sequential thoracoscopic approach with 4-5 ports on each side of the chest, sequential deflation and reinflation of the ipsilateral lung and significant potential for complications. The procedure is based on the “Dallas lesion set” concept ([2], [70]) creating RF lesions limited to the left atrium and superior vena cava (SVC). The entire procedure is confined within the boundaries of the pericardial space on the beating heart and could be achieved more simply from a single subxiphoid port with a snake robot with both high dexterity and stiffness; the goal is to replicate the function of the soft-tip endo-kitner retractor (labeled R) depicted in Figure 2: Five Box Maze and necessary for blunt dissection of the pericardial reflections and allow simultaneous exposure and ablation of target regions of the left atrium by delivery of RF energy. To perform the Five Box Maze, critical areas of the extrapericardial left atrium need to be dissected bluntly from the pericardial reflection and contiguous ablation lesions need to be created. The ability to perform a Dallas lesion set through a single port would amount to a breakthrough in the cure of AF.

A snake robot with high dexterity and stiffness will also allow new minimally invasive intracardiac procedures[68][69]. In addition, additional non-cardiac procedures in confined spaces where interactions with the tissue and the device are required will be enabled. For instance, video-assisted thoracic surgery (VATS) robotic lobectomy and mediastinoscopy[71] require complex dissection of the hilum of the lung and mediastinum. Trans-oral robotic surgery (TORS) will also benefit from such a hybrid snake robot . Successful completion of our proposed project will therefore introduce novel technologies that will be leveraged to advance other fields.

INNOVATION: To overcome the challenges of the proposed exemplary application, we plan to develop an easy-to-use novel image-guided hybrid snake robot that simultaneously provides tip dexterity and stiffness when it interacts with tissue, while all along providing the physician a real-time SOMETHINGthree-dimensional view of the target location from any perspective. We will achieve this ambitious goal through three aims:

Aim 1: We propose to combine, for the first time in this field, two types of snake robots: the HARP and a steerable cannula. The steerable cannula enhances the tip dexterity and allows reaching very tight spaces because of the small size. The HARP operates as stiff sheath for the steerable cannula so it significantly increases the tip stiffness of the steerable cannula. The hybrid snake promises the ability to follow a pre-planned arbitrary three dimensional curve and also the ability to perform dexterous contact tasks inside complex surgical environments.

Aim 2: We will also develop advanced path planners and tip controllers to exploit these unique abilities of the hybrid snake. Our path planners and tip controllers will allow safe navigation of the robot inside pericardium space while the robot tip can be in contact. The novelty of this approach will lie in our controller’s ability to modulating tip stiffness of the robot so we can set a low tip stiffness during navigating near delicate tissues and a high stiffness during blunt dissection. The planners will exploit this capability to coordinate both mechanisms, the HARP and the steerable cannula, as if they are one

Aim 3: We will design and implement a completely new method of image-guidance using electromagnetic position sensors on the robot projected over the left atrial surface mesh segmented from preoperative computerized tomography. This will allow safe and targeted performance of the task around the beating heart. This requires a new paradigm in image-guidance for MIS to obtain a more accurate and representative model of the operating site. In our opinion, the core enabling technology will be simultaneous localization and mapping (SLAM). SLAM is the task of having a robot build a map of its surrounding environment while simultaneously estimating its relative pose in that map. While SLAM has been investigated for wheeled mobile robots [56], we will be among the first to introduce and advance its explicit use for surgical robots. We propose SLAM for MIS that uses a stochastic filtering method, based on Kalman filtering, to continually correct surface models and estimate the motion of internal organs. Our SLAM algorithm will combine different modalities (vision, a magnetic tracking system and preoperative CT) to localize the robot as accurately as

Figure 2: Five Box Maze

possible. We believe the use of SLAM with surgery will ultimately improve the accuracy of navigation for MIS and will speed up operation times. This goes beyond the recent introduction of monte carlo localization techniques [58], in that we are performing motion estimation of the anatomical surfaces. We believe that the use of SLAM will allow MIS to be adopted by surgeons who normally resist the use of MIS. This is because SLAM will recover the “bird’s eye” view of the operating field to which surgeons have been accustomed. Finally, the SLAM algorithms developed in the proposed work are general to all procedures, and general to other robotic applications outside of the medical robotics field; this is truly a fundamental breakthrough.

We identified as the critical task blunt dissection, defined as a surgical technique used to gently separate tissues while avoiding injury to important nearby structures such as blood vessels or nerves. We believe that the hybrid snake will enable blunt dissection of pericardial reflections around the left atrium leading to a single port Dallas lesion set for cure of AF.

RESEARCH APPROACH: Progress Report (a) HARP Mechanism Development and Experiments. From June 2006 to present (October 2010), the Specific Aims of the previously awarded R01 grant were successfully achieved. We completed the development of the 11 mm Highly Articulated Robotic Probe (HARP) prototype for intrapericardial navigation and cardiac therapeutics delivery and technical details have been described elsewhere [62][75]. (Figure 3). A feeding mechanism, rigidly attached to the operating room table via an adjustable support mechanism, inserts the HARP through a small incision or port. The feeder supplies two additional degrees of freedom to regulate forward and reverse motion. In total, the HARP has 102 degrees of freedom

The HARP is sometimes called a “follow-the-leader” mechanism in that it can follow a tortuous path in three dimensions (Figure 4). It achieves this motion because of its unique concentric structure where an inner and outer mechanism alternate between being rigid and limp to generate a curve. Initially, the outer mechanism is made limp and advances forward. Meanwhile, the inner mechanism remains rigid and hence serves as a guide for outer mechanism. As the outer mechanism moves beyond the innte, it “steers” its distal portion. Subsequently, the outer mechanism is made rigid, the inner mechanism is made limp, the inner mechanism advances until it “catches up” to the outer mechanism, and the procedure repeats. The HARP truly is state of the art; it is innovative because it achieves the follow-the-leader motion with no exotic actuation technology, making it more maneuverable, robust, and reliable, all of which are important for carrying the concept to clinical use.

Finally, the HARP has three working channels through which conventional tools can be advanced as well as other novel devices such as the steerable cannulas (Figure 5). In our experiments, we have used these channels to house an ablation and mapping catheter up to 8 French in outside diameter, small fiberoptic cameras, biopsy forceps, and suction.

Figure 4: Probe leaving a feeder and then advancing around a heart model. This figure illustrates navigation around that heart which is not possible with a laparoscope or endoscope. The probe is an innovative design that uses conventional actuation technology, an innovative feature the NOTES robotics system will also have.

A B

D

ProbeFeeder

C

A BB

DD

ProbeFeeder

C

Figure 3: (a) HARP is mounted on a surgical table and a (c) feeder pushes the HARP out to follow (b,d) complex paths

We successfully tested the HARP in an anthropomorphic beating heart phantom, in a porcine model in vivo, and in human cadavers. We also tested and confirmed the hypothesis that the HARP is superior to existing dedicated rigid technology for subxiphoid videopericardioscopy [44][42][73]. In a recent study, we performed animal experiments comparing a commercially available subxiphoid videopericardioscopy system (SVP) (FlexView System, Guidant Cardiac Surgery) as a control (n=5) to the HARP in the closed-chest porcine preparation (n=5 in each group) [42] [35]. Multiple hemodynamic parameters were recorded during the testing. Unstable hemodynamics and fatal arrhythmia were observed during the trials using the rigid FlexView SVP device but we did not observe any occurrence of significant arrhythmia and unstable hemodynamics using HARP. The HARP was able to successfully navigate through the epicardial space of the beating heart from a single subxiphoid port to all 6 deep intrapericardial anatomical targets that were preselected via 7 distinct routes. Our study demonstrates the superiority of the HARP, with regard to navigation through the pericardial space to interventional targets on the epicardium, using image guidance alone. These findings suggest that such highly articulated systems may achieve similar improvements over rigid techniques with regard to patient safety during intrapericardial procedures in human patients.

(b) Kinematic Modeling and Control We have substantial experience on kinematic modeling, design and control of steerable cannulas [46][47][48]. We used Cosserat rod theory to model large deformation of the steerable cannula during contact. For robotic tasks that involve interaction with an environment, stiffness control should be superior to position control. In stiffness control, we are the first to establish a robust relation between the environment force and the tip of a continuum robot [46][47] This controller was demonstrated experimentally for a steerable cannula interacting with a soft environment. The steerable cannula consisted of two pre-curved NITi tubes of 15cm Figure 6a. An EM tracker was used to measure the tip position of the robot Figure 6b. Results showed that the stiffness controller can modulate the tip stiffness (Figure 6c.), provide good dynamic performance and exhibit stability during contact transitions[46].

(a) (b) (c)

Figure 6: Stiffness control of a steerable cannula (a) the tubes of the steerable cannula (b) the experimental setup wand EM tracker sensor (c) tip stiffness curves of steerable cannula for three different desired stiffnesses

(c) Identification of design constraints for Hybrid Snake Robot (unpublished preliminarily results)The hybrid snake we propose is composed of a HARP and a steerable cannula as shown in Error:

Reference source not found In our preliminary work, we identified the required strength for the HARP to accept a steerable cannula. The strength of the HARP should be high enough to be able to sustain the bending moment M caused by the steerable cannula. The bending moment caused by a straight steerable cannula

when it is deformed to the radius of curvature R is where is the Young's modulus of the

tubes of the steerable cannula, is the nominal diameter of the steerable cannula, and is the thickness of

Figure 5: HARP has three channels for passing tools

cannula. The maximum moment that the HARP can withstand is where is the cable tension, is the friction coefficient of the joint of the HARP and r is the radius of the HARP joint [62]. As a result, the HARP can sustain the cannula moment if

. We evaluated this equation for the current HARP and a steerable

cannula consisting of NiTi Tubes of 1.118 x 1.041 mm and 0.914 x 0.813mm Figure 7. The experimental results showed that the current HARP can deform the straight cannula to the radius of the curvature of 8.7 cm. Given the radius of curvature of the HARP is smaller than 8.7cm, this suggests that the HARP needs to be made stronger.

(d) Previous work for Simultaneous Localization and Mapping (SLAM): As a research group, we are building upon years of previous SLAM research to solve novel estimation problems. SLAM has largely been a navigation technology developed for mobile robots, and more recently free-flying rotor craft and AUVs [Ref from Howie Steve]. Our work in estimation and filtering has advanced the SLAM field in three key ways 1) robotic mapping with simple cameras (Tully 08), 2) mapping on a topological graph (Tully 09), and 3) localization for wheeled mobile robots (Tully 07). Our past experience in SLAM, combined with our expertise in snake robots makes us uniquely qualified to advance and extend SLAM to medical robotics. Already, we have begun to extend filtering and estimation to the HARP surgical robot. Our work, thus far, has involved performing state estimation and using stochastic models to estimate the most likely shape of the snake robot based on fusing information from the measured drive cable lengths and a position measurement from an EM tracker. We have also performed 3D reconstruction and registration to draw the snake robot location for image-guidance. This implementation has been used in our porcine preparation. An example result is shown in Figure 8. This preliminary work did not include the more advanced motion estimation that SLAM will provide.

SPECIFIC AIM #1: Design and develop a hybrid snake robot by integrating the HARP and the steerable cannula, via Sub-Aim #1.1: Develop a steerable cannula to be mounted inside the HARP to enable desired tissue dissection and ablation tasks Sub-Aim #1.2: Adapt the HARP’s unique link design and enhance its strength shape to accept the steerable cannula.

Figure 8: previous work on filtering for snake shape estimation and 3D reconstruction

Figure 7: Steerable cannula passing

through a working channel of the HARP.

Figure 11: Five box maze

Sub-Aim #1.1: Develop a steerable cannula to be mounted inside the articulated probe to enable desired tissue dissection tasks: We will design the steerable cannula of the hybrid snake such that it can perform two distinct tasks required for the Five Box Left Atrial Maze procedure [1]: 1) blunt dissection of pericardial reflections (Error: Reference source not found and Figure 10) and 2) ablation of the five box maze left atrial pattern (Figure 11). The four target pericardial reflections that require blunt dissection by the hybrid snake are (see Figure10): Blunt dissection of the four locations of pericardial reflections is required in order for the hybrid snake to be able to complete the maze ablation pattern. Successful dissection of reflection 1 will be defined by the ability of the

entire hybrid snake to advance through the dissected path. For reflections 2, 3 and 4 (see Figure 10), the success will be defined as the advancement of the tip of the steerable cannula past the anatomical boundaries of the reflection. We showed in the Table 1 the average distances between pulmonary veins, the IVC and the pulmonary veins and the diameters of the veins for five patients. We will consider these dimensions in designing our steerable cannula. For the blunt dissection, the steerable cannula should be able to make an up-and-down straight sweep motion when it pushes against the pericardium sack at each dissection location to

perform both ablation and dissection.

We will consider two preliminary designs for the steerable cannula. Our goal for the first design is to maximize the tip stiffness. High tip stiffness is required for performing blunt dissection and tissue retraction. Our goal for the second design is to maximize the tip dexterity by adding additional tip orientation control.

Design #1: This design will consist of two curved tubes of the same length, initial curve, and bending stiffness. The tubes will consist of two constant curvature segments: 1) a long straight segment which is longer than the length of the articulated probe and 2) a short curved section which is long enough to provide sufficient tip displacement for the intended task. The outer and inner diameters of both tubes are selected such that they

can be inserted inside each other and the combination can be inserted inside the port of the HARP. We refer to the combination of two tubes as combined tube. By rotating two tubes with respect to each other, the curvature of the curved segment of the combined tube varies from zero (straight configuration) to a maximum curvature. When the combined tube is inserted into the HARP, the straight portion takes the shape of the HARP. The combined tube can also extend and rotate inside the HARP and the orientation of its distal portion can be changed by the last link of the HARP.

Design #2. This design will consist of three curved tubes. Two of the tubes will be similar to the tubes of the first design. The third tube also consists of a straight segment and a curved segment. Comparing with two first tubes, the third tube has a longer straight segment, the radius of the curvature of its curved distal end is smaller, and its bending stiffness is much smaller. The third tube is inserted inside two first tubes. When the third tube is retracted, it conforms to the shape of the combination of the first two tubes. When it is extended, the curved segment relaxes to its original curvature. The second design provides additional tip orientation control and distal dexterity.

Diameter Average RangeLSPV 16mm 13-20mmLIPV 17mm 13-19mmRSPV 19mm 15-20mmRIPV 16mm 12-21mmDistance Average RangeLSPV-LIPV 5mm 3-8mmRSPV-RIPV 7mm 5-10mmIVC to RIPV 20mm 17-40mm

Table 1: measured value for 5 patients

Figure 12: The design 1 and 2 for the steerable cannula

Figure 10: Hybrid snake must dissect four spots of pericardium1) reflection between the right inferior pulmonary vein (RIPV) and the inferior vena cava (IVC); 2) reflection between the right superior pulmonary vein (RSPV) and the right pulmonary artery (RPA); 3) reflection between the left superior pulmonary vein (LSPV) and the left pulmonary artery (LPA) and 4- reflection over the superior vena cava.

Figure 9: Pericardial reflections (posterior view).

Areas in white indicate “extrapericardial” left atrium.

Both designs significantly increase the tip dexterity of the hybrid snake in comparison to the HARP alone. Figure 13: illustrating the curves and sweep tip motions achievable with the HARP alone vs. the hybrid snake. The hybrid snake (with tubes of 2mm) allows a longer tip displacement (8cm vs. 1cm) and a smaller radius of curvature (1cm vs. 4.5cm). The hybrid snake can go inside smaller spaces at its distal end (0.2cm vs. 1.2 cm).

Choice of material. We will construct our tubes from NiTi (Ni-Ti Tubes Inc, Fremont, CA). These tubes are super‐elastic and biocompatible. The minimum achievable radius of curvature for a NiTi tube is a function of tube outer diameter. The minimum achievable radius of curvature is 5mm for a tube of 1 mm diameter and 12.5 mm for a tube of 2.5mm. We will use heating and quenching process to manufacture our curved tubes [8]. Each tube is curved to desired shape and it is shape hold by a fixture. The fixture and the tubes are then heated in a high temperature oven and water quenched.

For intended tasks such as blunt dissection, retraction and ablation, we will modify the steerable cannula with addition of an atraumatic end (similar to endo-kitner,Figure 14) . For large tools (up to 8 French), we can advance them (e.g. radiofrequency ablation catheters or ICE) through the working port of the HARP and use the steerable cannula to navigate them. For smaller tools we can use the lumen of the steerable cannula itself.

Sub-Aim #1.2: Adapt the HARP’s unique link design and enhance its strength to accept steerable cannula: In our work during the previous NIH grant support period, the highly articulated probe had only been intended to carry highly flexible catheter-like devices. Because these types of catheter tools have been designed for specific applications (which are endoluminal in nature), they lack the qualities to make them useful for additional extended tasks in other environments. The introduction of steerable cannulas presents an interesting and necessary set of functionalities which stem from a combination of dexterity and strength. However, unlike highly flexible catheter devices, steerable cannulas provide a level of stability and strength orders of magnitude beyond their catheter-like counterparts. In our preliminary work (last section) we evaluated the bending moment induced by the steerable cannula on the HARP.

The HARP should be strong enough to sustain not only the bending moment induced by the steerable cannula but also the torque caused by the HARP weight and the tip load. The maximum moment caused by the HARP weight, W, and the external load F, is where L is the length of the HARP[62]. The

maximum moment that the HARP can generate is where is the cable tension, is the friction coefficient of the joint of the HARP and r is the radius of the HARP joint [62]. This proposal identifies three ways in which the HARP can be made stronger: material properties of the links, transmission cables and their driving motors, and increasing joint sizes. We will evaluate different materials for the links in the probe so that they can withstand higher compressive loads as well as improve the frictional characteristics which are critical to the performance of the HARP. While the specific material is yet to be determined, it is clear that a move to metal parts will improve intrinsic material strength as well as improve the coefficient of friction by 50% . Likewise, we will consider other cables but will start with braided stainless steel cable or Aramid cable because they allow a much larger transmission force. This also requires having larger motors or larger gear ratios. We will build a new HARP that has the required strength.

Finally, the hybrid snake will naturally require a new feeder. This portion of the work is straightforward engineering which will simply integrate the current HARP feeder with the feeder for a steerable cannula (Figure 15:). Our group is already well-poised to develop both the mechanism and control electronics architecture, leveraging our experience from the previous R01 support.

Figure 13: comparing distal dexterity of hybrid snake and HARP

Figure 14: endo-kitner

Figure 15: The feeder for the HARP and the feeder for a steerable cannula

SPECIFIC AIM #2 Develop control and path planning for the hybrid snake robot, via Sub-Aim #2.1: Kinematic modeling and position control, Sub-Aim #2.2: hybrid position-stiffness control and Sub-Aim #2.3: path planning.

A kinematic model is needed in order to precisely control the position or the stiffness of the tip of the hybrid snake. The kinematic model calculates the mapping between the tip configuration of the snake and the rotational angles of the motors of the snake. Without the kinematic model, it is impossible for the physician to precisely teleoperate the hybrid snake. We have developed separate kinematics models and controllers for the HARP [62][63] and the steerable cannula[46][47][48]. We will combine these models and controllers to control the hybrid snake. The overall kinematic model of the hybrid snake is decomposed into the product of the unloaded kinematic models of the HARP[62][63] and the steerable cannula[46][47][48], and a deformation model that calculates the bending of the hybrid snake due to the applied loading[46]. We will use the Cosserat theory for curved rods [93][94][46] to obtain the deformation of the snake. We extended the Cosserat rod theory to model large deformation of precuved tubes when they interact with each under external loading [46] [8][47][48].

A path planner is needed to coordinate the internal degrees of freedom of the hybrid snake to produce desired motions. Without a path planner, the physician would have to individually control each joint in the hybrid snake, which is an impossible, let alone unrealistic, task for the physician. Here, we will build upon the kinematic models developed in the previous subaims, to develop the theory, and then the software, which allows the physician to specifiy high-level directives and have the computer perform the necessary computations to carry out these objectives. This path planner will offset much of the cognitive load placed on the surgeon during a procedure allowing the physician to focus on important high-level decisions.

Sub-Aim #2.1: kinematic modeling and position control: We will develop a “two state” control for the hybrid snake. The controller at its “first state” will bring only the HARP to the targeted position and then at the second state will control the position of the tip of the steerable cannula. We will use a phantom haptic device (Sensible Inc. MA) to teleoperate the HARP using the “follow-the-leader” approach [62][63]. After the HARP reaches the intended configuration, we will lock the articulated probe at its position and then use the phantom device to control the position of the steerable cannula tip. We will use the unloaded kinematic model of the steerable cannula to control its tip. The unloaded kinematic model is obtained by using functional approximation of measured data [8][46] . Our main goal at this stage is to optimize the design of the hybrid snake, the bandwidth of its actuators and the accuracy of the unloaded kinematic model. We will test the hybrid snake and its controller on phantom animal models at the end of this sub-aim.

Sub-Aim #2.2: hybrid position-stiffness control: We will develop a hybrid position-stiffness controller to control both the tip stiffness and tip configuration of the hybrid snake. The hybrid stiffness controller will allow the surgeon to modulate the tip stiffness of the hybrid snake. The surgeon can set a high tip stiffness during blunt dissection and set a low stiffness when the snake tip approaches delicate tissues such as blood vessels. The stiffness controller will be very useful for performing blunt dissection of the pericardium reflections for the five box maze procedure. The physician can set a high tip stiffness when the tip of the snake is at the center of the marked dissection spots (Figure 10) and far from the pulmonary veins. The physician can set a low tip stiffness when the snake tip is close to the pulmonary veins. To implement the stiffness controller, we will use two EM trackers to measure the tip positions of the steerable cannula and the HARP. The outputs of these two sensors will then be used to calculate the deflection of the tip of the snake, the tip force, and the tip stiffness. If the measured tip stiffness of the hybrid snake is less than desired, the controller will advance the tip of the snake toward the tissue to increase the contact stiffness. If the measured stiffness is too high, the controller will move back the snake to reduce the applied force on the tissue and the tip stiffness.

We will use the hybrid stiffness controller to implement haptic feedback for the hybrid snake. In general, it is difficult to implement haptic feedback for a surgical robot due to the difficulty of using force sensor at the tip of the robot[14][74]. However, it is possible to measure the tip force of the hybrid snake from the deflection of its steerable cannula without using any force sensor. We will measure the deflection of the hybrid snake using the EM trackers and then will use the deformation model of the hybrid snake to calculate the tip force[46][47]. The calculated force then will be sent to the haptic device. It is expected that the haptic feedback will increase the accuracy of the physicians to perform blunt dissection with the snake[92].

Sub-Aim #2.3: Path Planning: Coordinating all of the internal degrees of freedom of the robot is an ambitious challenge, to say the least. Imagine trying to coordinate 102 knobs, one for each degree of freedom of the robot, to perform useful motion; it is impossible. Instead, the proposed work builds upon Choset’s background in path planning [67] to develop three modes of physician software support: 1) motion for the

articulated probe only, 2) motion for the steerable cannula, only, and 3) concurrent motion of the articulated probe and steerable cannula. Motion for the articulated probe is used for accessing the vicinity of the operative target; typically, the steerable cannula will be fully contained in the probe.

Given a desired path, the first mode of planning is seemingly relatively straightforward because the articulated probe is mechanically a follow-the-leader device, and was addressed in our previous work [62]. However, the surgeon may select a path that is not optimal for a particular task. For example, the surgeon may select a path that places the mechanism in an awkward configuration from which other motions are not possible or tip dexterity is compromised. To address this concern, we will develop a path optimizer that

maximizes the manipulability [34] of the mechanism along the path. Recall that manipulability is

where J is the Jacobian of the forward kinematic map derived in the previous subaim. By maximizing this quantity, one can maximize the set of velocities at the tip a mechanism can achieve, i.e., maximize its manipulability. The proposed work will consider many forms of manipulability, such as manipulability integrated along the path, manipulability integrated, yet weighted by arc length, along the path, etc. This type of optimization has not been done for articulated robots.

Path planning for the steerable cannula: The significant contribution to path planning happens when we consider simultaneous path planning for the articulated probe and the steerable cannula. The question becomes: which device should be moved to achieve a desired tip motion. Once again, we take recourse to notions of manipulability to maximize the dexterity of the overall system. Let Jp and Js be the Jacobians of the probe and steerable cannula, respectively. Now, form a new Jacobian by stacking these two matrices on top of each other. This meta-Jacobian can then be used in the manipulability calculation, resulting in a measure of tip dexterity for the hybrid snake. Now, one can specify an optimization problem that maximizes overall manipulability to move the tip along desired directions. Such an optimization process will yield a planner that optimally moves the probe and cannula. This portion of the proposed work will investigate variations of this measure and validate this approach for planning.

We should also stress that often the hybrid snake may be in contact with the environment, and it may be necessary to plan paths that regulate the force interaction between the mechanism and the anatomic target. To

plan accordingly, we make use of force-manipulability which is . Extremizing this quantity allows

the device to have a wide range of force control over the tip, but naturally at the loss of variations in velocity. This makes sense in that power, the product of force and velocity, must be conserved. The proposed work will consider variations of these measure that combine manipulability and force-manipulabilty, say linear combinations of the two. Such a planner has not been considered on highly articulated snake devices before, and potentially represents a significant contribution to the path planning literature.

Finally, developments in path planning will assist the surgeon when performing the blunt dissection process. Using the contributions to planning, described above, we will standardize the motion of the steerable cannula: into an up-and-down sweep followed by a right-to-left sweep, each repeated 3 times. This type of motion is called coverage (Choset, 2000, Acar, 2002), a type of planner that passes a robot end-effector over all points in a target region. Our research group lifted coverage into non-flat surfaces in three dimensions in (Atkar 2005); this work also considered optimizing a process variable, such a uniformity of paint in a surface deposition application. In the proposed work, we will build upon the prior coverage work to optimize deviation from a desired force profile while covering a non-flat surface, using the force path planners described above. With this planner in-hand, the surgeon can specify a region over which the two pairs of three paths should be followed, and then will produce a path that covers the region while maintaining a desired force profile.

Specific Aim #3: Develop 4D 3D image-guidance that shows the shape and position of the snake robot registered to a patient-specific cardiac reconstruction. Sub-Aim #3.1: CT data collection and segmentation; Sub-Aim #3.2: Online reconstruction and SLAM; Sub-Aim #3.3: 34D rendered graphical interface.

Sub-Aim #3.1: CT data collection and segmentation: Preoperative imaging is usually performed for diagnostic purposes, but can also be useful for preoperative planning via the reconstruction of 3D surfaces from 2D image slices. These 3D reconstructions are also beneficial for live image-guidance when registered to intraoperative information such as real-time ultrasound or an electromagnetic tracker. 3D surface reconstruction is obtained by segmenting the 2D image slices so as to associate each pixel with a specific organ or internal structure. This process can be performed by thresholding the pixel values, by using the marching cubes algorithm [59], by manually segmenting portions of the images, or by implementing any

combination of the three aforementioned methods. We will use the open-source multi-platform 3D Slicer [61] software package to perform 3D anatomical reconstruction from CT data.

When we scan a subject using CT angiography, the output will be a set of DICOM files that are easily imported into the 3D Slicer software package. From here, we will segment the images and create a 3D reconstruction of the epicardial surface (as well as additional structures) using conventional segmentation techniques. By implementing an extra C++ module, we will have Slicer output a custom data file that stores the surfaces as separate triangle meshes that are compressed at different levels of detail and that are in a format that is easily parsed by our SLAM software as an initial reconstruction for initializing our 34D SLAM algorithm.

In initial testing, we collected CT data for five patients. The CT data shows that the pericardial reflections can be seen by CT imaging. We will develop an additional C++ module for Slicer that will automate the segmentation of the pericardial reflections.

Sub-Aim #3.2: Online reconstruction and SLAM: A static 3D reconstruction from CT images is not a representative model of the operating site. We will use SLAM to estimate patient-specific motion parameters of the heart tissue in order to create a more informative 34D model. Our SLAM algorithm will also provide accurate estimation of the relative pose of the snake robot, further enhancing the physician’s situational awareness. This implies the need for accurate registration to the 34D reconstruction.

Our SLAM method is initialized from the 3D static epicardial surface reconstruction from Sub-Aim 3.1. We will essentially use the static surface as a crude initial guess for the true 34D surface model. At first, this model will be incorrect, but through the SLAM algorithm’s filtering process, the motion parameters for each element of the surface triangle mesh will be incrementally adjusted when new intraoperative information is available.

One type of measurement that will help determine motion of the heart tissue, and which will feed directly into the Kalman filtering framework that we will adopt for SLAM, is detected cyclical motion of the EM tracker at the distal end of the snake robot. We will set the tip stiffness of the hybrid snake to a low value and navigate the tip of the snake along the surface of the heart. The position data will provide a means to correct, online, the surface parameters of the triangle mesh. We note that we will also navigate the hybrid snake through the pericardial reflections to further refine surface and motion information of this anatomical target.

Another measurement that we will rely upon for estimating the motion parameters of our updated 34D model will be visually tracked features from camera images that are recorded at the distal end of the snake. The measurements we obtain from this method are known in the robotics community to be “bearing-only”, which unfortunately requires unique attention for SLAM problems. Luckily, we already have in depth experience performing a similar visual feature mapping task for SLAM [CITE].

While the SLAM filter itself will be based on Kalman filtering, a widely used estimation technique for linear problems, we handle the nonlinearity inherent in the kinematic models for the hybrid robot by developing and then implementing a novel iterative filtering technique to achieve better performance for motion estimation. This iterative method will be based on proven numerical optimization algorithms. Additionally, our SLAM algorithm will use stochastic models, and thus, will maintain error bounds on the motion parameters that we are estimating for the heart surface.

Lastly, the SLAM algorithm will automatically generate, via filtering and parameter estimation, the most likely pose of the robot. This means that the registration of the robot to the CT images will be updated and corrected automatically during a surgical trial or an experiment. The initial guess for the registration parameters, which will be used to initialize SLAM, will be based on the matching of 9 externally applied fiducial markers that are visible in the CT images.

Sub-Aim #3.3: 3 4 D Rendered Graphical Interface: For image-guidance, we propose an intuitive graphical interface that will display, on a computer screen, a

34D rendered visualization of the operation which allow the physician to view the operating site from any vantage point. This graphical interface will involve the incorporation of new software modules into the 3D Slicer software package [CITE]. The first step will be to implement a custom module that will render the snake robot among the surface models of the heart. Both the surface models and the snake position estimate will be fed from an independently running software process that is performing the SLAM algorithm discussed in Sub-Aim 3.2. Therefore, a second software module will be developed to load and parse SLAM data into Slicer from this secondary process.

The SLAM algorithm produces a 34D model, and thus it is important that we synchronize the animationthe update of the rendered epicardial surface with other the cardiac parameters, such as stiffness,motion of the subject. This will be achieved by inputting a live ECG signal from the subject into the computer that is running

the graphical interface. The ECG will be used for timing purposes so that the appropriate model is rendered at the correct instance in the cardiac cycle.

One focus of this graphical interface will be to make the software as intuitive as possible for both surgeons and engineers. This means that the inner workings and technical details of our SLAM algorithm will be hidden “under the hood”. Only the most likely rendered view of the operation will be displayed at any given time, which will simplify image-guidance for the user. If needed, the software will allow for easy switching between the rendered visualization, direct vision, and/or a view of the CT DICOM files for planning. To show the direct camera feed, we will need to write additional C++ code that will also interface with Slicer and that will process the image data. Implementing this intuitive interface will involve the development of C++ software for system integration.

Potential Pitfalls and Solutions: There are several challenges associated with the development of the hybrid snake and its controllers. The first concern with the hybrid snake is the potential complexity due to the length of its steerable cannula. As the length of the steerable cannula increases, the torsional twist of the concentric tubes will increase and other potential phenomena may affect the kinematic mapping of the steerable cannula. The second concern is the required strength for the articulated probe to bend the steerable cannula. To address these concerns, we will have the option of using multi-segment pre-curved tubes that each segment has different material properties. We will consider using a material with low bending stiffness and high torsional stiffness for portions of the tubes which are inside the articulated probe and use a material with a high bending stiffness for the portion which is outside of the articulated probe. We will also use functional approximation for building kinematics mapping that allows computationally expensive model to be implemented.

The third concern is the instability and oscillation that may cause due to the use of tip sensors in closed loop controllers. In order to prevent possible instability and oscillations, we will use aggressive analytical methods based on passivity theory and absolute stability theory to ensure the stability of hybrid snake in worst case scenarios. Our team is uniquely qualified to address all of these issues.

Finally, potential pitfalls may occur while running the SLAM algorithm for image guidance. If we happen to miss-associate measurements from camera images with the wrong moving features in the 4D surface estimated by SLAM, then the estimated surface will have positional error. With robust visual feature tracking and conservative estimation, though, we believe this situation would be a rare occurrence. Also, if metal objects influence the accuracy of the EM tracker due to disruption to the generated magnetic field, both the snake robot and surface estimation within the SLAM filtering algorithm may be negatively affected. Any additional error like this will cause the 34D image-guidance rendering to be a misrepresentation of the operating site, which is cause for concern. In our preliminary work, though, we have thoroughly tested the EM tracker inside and outside of a surgical environment and are confident in its accuracy when integrated within the HARP robot.

SPECIFIC EXPERIMENTAL PROTOCOLS: We will use our well established porcine preparation for all animal experiments, starting in Year 2 (see Table X). Large (35-45kg) healthy swine of either sex will be received by the Harvard Medical School Animal Research Facility (West Roxbury VABHCS Campus) and housed at least overnight following delivery.

On the first experimental day, subjects will receive intramuscular injections of 20 mg/kg ketamine and 2 mg/kg xylazine and 1% to 5% isoflurane will be delivered using a face mask for general endotracheal anesthesia (GETA). Nine metal fiducial markers will be placed on the skin of the ventral thorax and computed tomography (CT) will be performed. 3D CT images will be obtained using an helical CT scanner (64-slice LightSpeed VCT; GE Healthcare, Milwaukee, WI). CT scanning (120kV, 800mA, pitch of 0.16:1, 350ms/rotation gantry speed) with a thickness of 0.6mm after intravenous injection of an iopamidol contrast agent. Following the procedure, subjects will be extubated and observed in the recovery room. The CT DICOM file will be obtained from the Radiology Department and volume rendering with SLICER or other software will be performed.

On the second experimental day, subjects will receive GETA and will be placed supine on the operating table. Heart rate and rhythm will be monitored continuously with electrocardiography. The right carotid artery and jugular vein will be exposed through an incision on the right side of the neck and the right carotid artery will be cannulated with a 6 French catheter to monitor the arterial blood pressure. The jugular vein will be cannulated with a 7 French Swan-Ganz catheter to monitor the central venous pressure, the

pulmonary artery pressure and mixed venous oxygen saturation derived from blood samples obtained from the tip of the Swan-Ganz catheter placed in the pulmonary artery.

The AURORA tracking system (NDI, Waterloo, Canada) will be mounted on the operating table. The working tip of the hybrid snake is tracked with respect to these fiducials using an embedded three-axis magnetic tracking coil. A user interface will graphically represent the heart and surrounding mediastinal structures.

The hybrid snake will be introduced inside the pericardial cavity through a single 15mm Subxiphoid port as previously described [37].The HARP will be advanced using both direct endoscopic visual feedback and image guidance with the display interface. Once the HARP reaches the target reflection, it will be locked into place and the steerable cannula will be introduced and advanced (MM: alternative is having the HYBRID SNAKE already assembled). Additional imaging will be acquired with an 8FR AcuNav intracardiac ultrasound probe (ICE, Acuson Inc) in order to identify the RIPV and IVC and measure the distance. Blunt dissection of the target pericardial reflection will be obtained by repeating a gentle alternating sweeping motion of the steerable cannula according to a plan taking into consideration the dexterity of the hybrid snake at the target location and the intravascular and intracardiac pressure measured with ICE. The path planning technology, described in Aim 2.3, will reduce the cognitive load on the surgeon to achieve blunt dissection: instead of directly controlling many degrees of freedom, the surgeon specifies a region in an image of the operating site, and then the planner automatically the steerable cannula in an up-and-down sweep, repeated 3 times, followed by a right-to-left sweep, also repeated 3 times. The range of the sweep will be predetermined based on the combined ICE and CT measurement (e.g. distance from the cranial edge of the inferior vena cava to the caudal edge of the right inferior pulmonary vein for target #1).

Successful dissection of reflection #1 will be defined by the ability of the entire hybrid snake to advance through the dissected path. For reflections #2, 3 and 4, success will be defined as the advancement of the tip of the steerable cannula past the anatomical boundaries of the reflection. Safety will be monitored by absence of blood in the pericardial cavity under direct video monitoring, lack of hemodynamic changes and arrhythmias. Following the completion of all 4 blunt dissections, a sternotomy will be peformed and each dissection will be tested with the Wolf Lumitip Dissector (Atricure, West Chester, OH).

For each blunt dissection, we will record the following parameters: (a) time from first steerable cannula dissection to confirmation of successful completion of task; (b) extent of deformation of surrounding vascular or cardiac structures (in mm) by ICE; (c) subjective level of difficulty by the operator on a scale of 1 (very difficult) to 5 (very easy); (d) presence or absence of complications.

Following complete dissection of the 4 pericardial reflections, a commercially available RF ablation catheter (ThermoCool, Biosense-Webster) will be advanced through a working port and, using a combination of imaging including direct video endoscopy and ICE, the catheter will be used to perform the epicardial five box Maze. The steerable cannula of the hybrid snake will be used to facilitate access of the RF catheter by providing retraction of atrial tissue. Bidirectional (i.e. entry and exit) block will be confirmed for each one of the five boxes, and additional ablation will be used until complete block is confirmed.

We propose to compare the HARP to the hybrid snake for the ability to a) complete the dissection of the 4 pericardial reflection targets and (b) perform the five box Maze.

Data Analysis: Quantitative results will be expressed as the mean percent change from each subject baseline of a given parameter ± the standard error of the mean (SEM) and analyzed using a software package for statistical analysis (Stata/IC software, version 10.1; Stata Corporation, College Station, TX). These percentage changes will be compared between groups. Wilcoxon’s signed-rank test will be used to determine the significance of the difference. P value of less than 0.05 will be considered statistically significant. Arrhythmogenicity will be reported qualitatively in a binary manner as either the presence or absence of operator-identifiable arrhythmia on EKG.

Work Plan: The work plan spans five years, and the specific aims and tasks will be completed as shown in the table and chart the below.

YEAR

Investigators

Choset has been leading a research group for over 15 years developing a comprehensive program on snake robots. These work touches about key areas: mechanism design, path planning, control, and estimation. Choset’s research group has developed the most varied collection of snake robots in the nation, and these devices have been patented, as well as deployed for real use. Choset goes beyond building the device to enabling it to perform purposeful motion. Choset has received many awards, such as the NSF Career Award and ONR Young investigator award, for his work on path planning; five years ago, he published a book path planning with the MIT Press where Choset is the first author. Choset’s students have won many best paper and video awards on their work on control, most recently best video for a climbing robot at the International Conference on Robotics and Automation. Finally, Choset’s group has developed estimators to localize mobile robots in large spaces. Few researchers in robotics work on all four of these topics and no other group can integrate them.

Harvard Medical School

Carnegie Mellon University

Year 1 Design 1 of steerable cannula (Aim 1) Feeder design (Aim 1)CT data and collection (Aim 3)

Kinematic model & position control (Aim2)Year 2 Test snake in phantom (Aim1-2)

Design 2 of steerable cannula & control (Aim 1-2)Design a strong probe (Aim 1)Construct new feeder (Aim1)Path planning (isolated mechanisms) (Aim 2)Online reconstruction and SLAM (Aim 3)

Preliminary animal test (N=5)Year 3 Stiffness control of snake [Aim 2]

Animal testing for pericardial dissection (N=10) Continue on design of the probe (Aim1)Path planning (combined mechanism) (Aim2)34D rendered and SLAM (Aim 3)

Year 4 User interface design for snake [Aim 3] Imaging & segmentation[Aim 3]Animal testing (N=15) [Aim 1-3]

Iterate on feeder design (Aim 1)Path planning using results from SLAM (Aim2)34D Rendering graphical interface (Aim 3)

Year 5 Imaging & segmentation [Aim 3]Snake vs HARP for dissection (N=20) Five Box Ablation in human cadavers (N=5)

Fully integrate interface, planning and SLAM (Aim 2 and 3)

References[1] Sirak J, Jones D, Schwartzman D. The Five-Box Thoracoscopic Maze Procedure. Ann Thorac Surg

2010;90:986–90[2] JL Cox - The longstanding, persistent confusion surrounding surgery for atrial fibrillation,The Journal of

Thoracic and Cardiovascular Surgery, 2010 [3] Robinson R, Davies J. Continuum robots - a state of the art. in IEEE International Conference on

Robotics and Automation, vol. 4, 1999,, pp. 2849–2854.[4] Camarillo D, Milne C , Carlson C, Salisbury KJ, “Mechanics modeling of tendon-driven continuum

manipulators,” IEEE Transactions Robotics, vol. 24(6), pp. 1262–1273, 2008.[5] Jones B, Walker I, “Kinematics for multi-section continuum robots,” IEEE Transactions on Robotics, vol.

22, no. 1, pp. 45–53, 2006.[6] Simaan N, Taylor R, Flint P. A dexterous system for laryngeal surgery. in Proc. IEEE Intl. Conf. on

Robotics and Automation, New Orleans, April 2004, pp. 351–357.[7] R. Webster, J. Romano, and N. Cowan, “Mechanics of precurved-tube continuum robots,” IEEE

Transactions Robotics, vol. 25(1), pp. 67–78, 2009.[8] P. Dupont, J. Lock, B. Itkowitz, and E. Butler, “Design and control of concentric tube robots,” IEEE

Transactions Robotics, vol. 26(2), pp. 209–225, 2010.[9] P. Sears and P. Dupont, “A steerable needle technology using curved concentric tubes,” in IEEE/RSJ

Int. Conference on Intelligent Robots and Systems, Beijing, 2006, pp. 2850–2856.[10] Http://www.hansenmedical.com/ .[11] R. J. Webster, J. S. Kim, N. J. Cowan, G. S. Chirikjian, and A. M. Okamura, “Nonholonomic modeling

of needle steering, ”International Journal of Robotics Research, vol. 25, no. 5–6, pp. 509–525, 2006.[12] N. Hogan, “Impedance control: An approach to manipulation: Part i, ii, iii,” J. DYN. SYST. MEAS.

CONTROL, vol. 107, no. 1, pp. 1–24, 1985.[13] J. K. Salisbury, “Active stiffness control of a manipulator in Cartesian coordinates,” in Proc. IEEE Conf.

Decision Control, 1980, pp. 95 –100.[14] M. Mahvash and A. Okamura, “Friction compensation for enhancing transparency of a teleoperator

with compliant transmission,” IEEE Transactions on Robotics and Automation, vol. 23, no. 6, pp. 1240–1246, 2007.

[15] K. Xu and N. Simaan, “An investigation of the intrinsic force sensing capabilities of continuum robots,” IEEE Transactions on Robotics, vol. 24, no. 3, pp. 576–587, 2008.

[16] D. Pai, “Strands: Interactive simulation of thin solids using cosserat models,” in Proceedings of Eurographics’02, vol. 21, no. 3, 1990, pp. 347–352.

[17] S. S. Antman, Nonlinear problems of elasticity. Springer Verlag, NewYork, 1995.

[18] J. Lock, G. Laing, M. Mahvash, and P. Dupont, “Quasistatic modeling of concentric tube robots with external loads,” in IEEE/RSJ Int. Conference on Intelligent Robots and Systems, in press, 2010.

[19] D. B. Camarillo, K. Loewke, C. Carlson, and K. J. Salisbury, “Vision based 3-d shape sensing of flexible manipulators,” in IEEE International Conference on Robotics and Automation, Pasadena,CA, 2008, pp. 2940–2947.

[20] K. Xu and N. Simaan, “Analytic formulation for kinematics, statics and shape restoration of multi-backbone continuum robots via elliptic integrals,” ASME Journal of Mechanisms and Robotics, vol. 2, no. 1, pp. 1–13, 2010.

[21] C. Chiou and M. Shahinpoor, “Dynamic stability analysis of a two linkforce-controlled flexible manipulator,” ASME J. Dyn. Sys.

Meas., vol. 112, no. 2, pp. 661–666, 1990[22] A. Moglia, A. Menciassi, M. Schurr, annd P. Dario, “Wireless capsule endoscopy: From diagnostic

devices to multipurpose robotic systems,”Biomed. Microdevices, vol. 9, no. 2, pp. 235–243, Apr. 2007.[23] Patronik NA, Ota T, Zenati MA, Riviere CN. A miniature mobile robot for precise and stable access to

the beating heart. IEEE Transactions on Robotics, Issue Date: vol 25 (5) pp: 1109 – 1124, Oct. 2009. [24] R. J. Webster III, J. S. Kim, N. J. Cowan, G. S. Chirikjian and A. M. Okamura, "Nonholonomic Modeling

of Needle Steering," International Journal of Robotics Research, Vol. 25, No. 5-6, pp. 509-525, May-June 2006

[25] S. P. DiMaio, S. E. Salcudean, “Needle insertion modeling and simulation,” IEEE Trans. Rob. Autom., vol. 19, no. 5, pp. 864-875,October 2003.

[26] S. Okazawa, R. Ebrahimi, J. Chuang, S. E. Salcudean, R. Rohling, “Hand-held steerable needle device,” IEEE ASME Trans. Mechatron.vol. 10, no. 3, June 2005.

[27] J. A. Engh, G. Podnar, D. Kondziolka, C. Riviere, “Toward effective needle steering in brain tissue,” in Proc. 28th Annual Int. Conf. IEEE Eng. Med. Biol. Soc., 2006, pp. 559-562.

[28] H. B. Brown, M. Schwerin, E. Shammas, and H. Choset, "Design and control of a second-generation hyper-redundant mechanism," in Proc. IEEE/RSJ Int. Conf. Intelligent Robots and Systems, 2007, pp. 2603-2608.

[29] Bruno Siciliano , Oussama Khatib , “Handbook of robotics” Page 347, Springer, 2008

[30] P. Atkar, A. Greenfield, D. Conner, H. Choset and A. Rizzi, “Uniform Coverage of Simple Automotive Surfaces,” International Journal of Robotics Research, Vol. 21, pp. 883-898, 2005

[31] Mingyen Ho, and Jaydev P. Desai, “Characterization of SMA actuator for application in robotic neurosurgery”, in International Conference of the IEEE Engineering in Medicine and Biology, pp 6856-6859, September, 2009

[32] E. Acar, H. Choset, A. Rizzi, P. Atkar, and D. Hull. “Morse Decompositions for Coverage Tasks,” International Journal of Robotics Research, Vol. 21, No. 4, pp. 331-344, April, 2002.

[33] H. Choset, Autonomous Robots, “Coverage of Known Spaces: The Boustrophedon Cellular Decomposition,” Vol. 9, pp. 247-253, 2000

[34] Tsuneo Yoshikawa. Manipulability of Robotic Mechanisms. The International Journal ofRobotics Research, 4(3), 1985.

[35] Yokota T, Ota T, Schwartzman D, Zenati MA. Impact of Subxiphoid Video Pericardioscopy with a Rigid Shaft on Cardiac Hemodynamics in a Porcine Model. Innovations 2010;5(1):51-54. PMCID2854418

[36] Yokota T, Ota T, Schwartzman D, Nikou C, Jaramaz B, Choset H, Zenati MA. A Highly Articulated Robotic Surgical System using an External Visualization and Navigation System. Int J CARS 2009;4(Suppl.I):S245-S303 [Abstract]

[37] Zenati MA, Yokota T, Schwartzman D, Ota T, Zubiate B, Wright C, Nikou C, Choset H. Model-Guided Epicardial Interventions using a Novel “Snake” Robot (CARDIO-ARM) through a Subxiphoid Port Access. Proceedings, The 14th International IASTED Conference on Robotics and Applications, 2009;242-246 NIHMSID148983

[38] Yokota T, Ota T, Schwartzman D, Nikou C, Jaramaz B, Zubiate B, Choset H, Zenati MA. Model-Guided Single Port Interventions Using a Novel Highly Articulated Mechatronic System. Proceedings, Annual Meeting of the Biomedical Engineering Society (BMES), Pittsburgh, PA, October 7-10, 2009, page 38. [Abstract]

[39] Choset H, Zenati MA. Enabling Medical Robotics for the Next Generation of Minimally Invasive Procedures. Proceedings, IEEE/ICRA, Anchorage, Alaska, May 2010 [Abstract]

[40] Yokota T, Ota T, Schwartzman D, Nikou C, Zubiate B, Choset H, Zenati MA. Model-Guided Single Port Epicardial Interventions using a Novel Highly Articulated Mechatronic System. Proceedings, 1st Worldwide Congress of the Clinical Robotic Surgery Association (CRSA), Chicago, IL, October 9-10, 2009, page 19 [Abstract].

[41] Yokota T, Ota T, Zubiate B, Jaramaz B, Zenati MA. Model-guided Single Port Epicardial interventions using a Highly Articulated Robotic System. Proceedings, Minimally Invasive Robotic Association (MIRA) 2010 Meeting, San Diego, CA, January 28-20, 2010. Page 46 (S080) [Abstract].

[42] Chapman MP, Yokota T, Ota T, Tully S, Schwartzman D, Choset H, Zenati MA. A Highly Articulated Robotic System (CardioARM) is Safer than a Rigid System for Intrapericardial Intervention in a Porcine Model. Proceedings, IEEE International Conference on Robotics & Automation (ICRA), Workshop on “Snakes, Worms and Catheters: Continuum and Serpentine Robots for Minimally Invasive Surgery”, Anchorage, Alaska, May 3, 2010 PMCID187594

[43] E. Paljug, T. Ohm, S. Hayati, “The JPL Serpentine Robot: a 12 DOF System for Inspection,” IEEE Conf. on Robotics and Automation,pp. 3143-3148, 1995

[44] Zenati MA, Yokota T, Ota T, Choset H, Schwartzman D. Improved Hemodynamics for Single-Port Subxiphoid Intrapericardial Navigation with a Highly Articulated Robotic System (CardioARM) vs. a Rigid Shaft Device. Proceedings, International Society for Minimally Invasive Cardiac Surgery (ISMICS), 2010 Annual Meeting, Berlin, Germany, June 2010 [Abstract].

[45] http://www.cambridgeendo.com/ [46] M. Mahvash, and P. E. Dupont, Stiffness control of surgical continuum manipulators, IEEE

Transactions on Robotics, (accepted). [47] M. Mahvash, and P. Dupont, Stiffness Control of a Continuum Manipulator in Contact with a Soft

Environment, IEEE /RSJ International Conference on Intelligent Robots and Systems (IROS), 2010 (in press)

[48] 3- J. Lock, G. Laing, M. Mahvash, and P. Dupont, Kinematic Modeling of Concentric Tube Robots Incorporating Externally Applied Loading , IEEE /RSJ International Conference on Intelligent Robots and Systems(IROS), 2010 (in press).

[49] Berlinger NT. Robotic surgery—squeezing into tight places. New Engl J Med. 2006;354:2099–2101.[50] Camarillo DB, Krummel TM, Salisbury JK. Robotic technology in surgery: Past, present, and future

American Journal of Surgery - Volume 188, Issue 4 (October 2004)  - 2004 Elsevier . [51] RH Taylor, D Stoianovici, Medical robotics in computer-integrated surgery, IEEE transactions on

robotics, and automation, vol. 19, no+. 5, oct 2003 [52] http://gizmodo.com/338982/ “I snake robot to offer slithering assistance during surgery”[53] Yilmaz A, Geuzebroek GSC, Van Putte BP, Boersma LVA, Sonker U, De Bakker JMT, Van Boven WJ.

Completely thoracoscopic pulmonary vein isolation with ganglionic plexus ablation and left atrial appendage amputation for treatment of atrial fibrillation. Eur J Cardiothorac Surg 2010;38:356-360

[54] Edgerton JR, Jackman WM, Mack MJ. A new epicardial lesion set for minimal access left atrial maze: the Dallas lesion set. Ann Thorac Surg 2009;88:1655-7.

[55] D’Avila A, Scanavacca M, Sosa E, Riskin JN, Reddy VY. Pericardial anatomy for the Interventional Electrophysiologist. J Cardiovasc Electrophysiol 2003;14:422-430

[56] M. Dissanayake, P. Newman, H. Durrant-Whyte, S. Clark, and M. Csorba. A solution to the simultaneous localisation and map building (SLAM) problem. IEEE Transactions of Robotics and Automation, 17(3):229–241, June 2001.

[57] M.W. Vannier, J.L. Marsh, and J.O. Warren. Three dimensional CT reconstruction images for craniofacial surgical planning and evaluation. Radiology. 1984.

[58] A.B. Koolwal, F. Barbagli, C. Carlson, D. Liang. An Ultrasound-based Localization Algorithm for Catheter Ablation Guidance in the Left Atrium. The International Journal of Robotics Research. 29: 643-665. May 2010.

[59] W. Lorensen, and H.E. Cline. Marching Cubes: A High Resolution 3D Surface Construction Algorithm. In Computer Graphics, Vol. 21, No. 4, July, 1987, pp. 163-169.

[60] L. Piegl. On Nurbs: a survey. IEEE Computer Graphics and Applications. 11(1):55-71. January, 1991.

[61] D.T. Gering, A. Nabavi, R. Kikinis, N. Hata, L.J. O’Donnell, W.E. Grimson, F.A. Jolesz, P.M. Black, and W.M. Wells. An integrated visualization system for surgical planning and guidance using image fusion and an open MR. Journal Magnetic Resonance Imaging.2001.

[62] A. Degani, H. Choset, A. Wolf, and M. Zenati, Highly Articulated Robotic Probe for Minimally Invasive Surgery, Proc. 2006 IEEE International Conference on Robotics & Automation, Orlando, USA, May 2006

[63] H. Choset and W. Henning. A follow-the-leader approach to serpentine robot motionplanning. ASCE Journal of Aerospace Engineering, 12(2):65{73, April 1999

[64] Iterated Filters for Bearing-Only SLAM. Stephen Tully, Hyungpil Moon, George Kantor, and Howie Choset . Proc. 2008 IEEE International Conference on Robotics & Automation, Pasadena, California, USA, May 2008.

[65] A Multi-Hypothesis Topological SLAM Approach for Loop Closing on Edge-Ordered Graphs. Stephen Tully, George Kantor, Howie Choset, and Felix Werner. Proc. 2009 IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems, Saint Louis, MO, USA, October 2009.

[66] Hybrid Localization Using the Hierarchical Atlas. Stephen Tully, Hyungpil Moon, Deryck Morales, George Kantor, and Howie Choset. In Proceedings of the 2007 IEEE/RSJ Intl. Conference on Intelligent Robots and Systems, San Diego, CA, USA, Oct 29 - Nov 2, 2007, Pages 2857-2864.

[67] H. Choset, K. Lynch, S. Hutchinson, G. Kantor, W. Burgard L. Kavraki, S. Thrun, Principles of Motion Planning, MIT Press, May, 2005.

[68] Block PC. Percutaneous Mitral Valve Repair for Mitral Regurgitation. J Interv Cardiol. 16(1):93‐96, 2003.

[69] Y Suematsu, PJ del Nido - Robotic pediatric cardiac surgery: present and future perspectives, The American Journal of Surgery, 2004 – Elsevier

[70] Z. J.   Edgerton , J. R.   Edgerton , History of surgery for atrial fibrillation, Heart Rhythm, Elsevier, Vol 6-12, December 2009.

[71] F Gharagozloo, F Najam, Robotic surgery- McGraw-Hill Medical, 2009

[72] Ota T, Degani A, Zubiate B, Wolf A, Choset H, Zenati MA. Epicardial Atrial Ablation using a Novel Articulated Robotic Medical Probe via a Percutaneous Subxiphoid Approach. Innovations 2006;1(6):335-340. PMCID: PMC1993852

[73] Ota T, Degani A, Schwartzman D, Zubiate B, McGarvey J, Choset H, Zenati MA. A Highly Articulated Robotic Surgical System for Minimally Invasive Surgery. Ann Thorac Surg 2009;87:1253-6. PMCID: PMC2691642

[74] M. Mahvash, J. Gwilliam, R. Agarwal, B. Vagvolgyi, L. Su, D. Yuh, and A. Okamura, Force feedback surgical teleoperator: controller design and palpation experiments, Symposium on HapticInterfaces for Virtual Environments and Teleoperator Systems, 2008

[75] Degani A, Choset H, Wolf A, Ota T, Zenati MA. Percutaneous Intrapericardial Interventions Using a Highly Articulated Robotic Probe. Proceedings, First IEEE/EBIB International Conference on Biomedical Robotics and Biomechatronics (BioRob’06) 2006, pp. 7-12 DOI 10.1109/BIOROB.2006.1639051 (full length peer-reviewed manuscript)

[76] Ota T, Degani A, Schwartzman D, Zubiate B, McGarvey J, Choset H, Zenati MA. A Novel Highly Articulated Robotic Surgical System for Epicardial Ablation. Conf Proc IEEE Eng Med Biol Soc 2008;1:250-53 (full length peer-reviewed proceedings manuscript). PMCID: PMC2630583

[77] Degani A, Choset H, Zubiate B, Ota T, Zenati MA. Highly Articulated Robotic Probe for Minimally Invasive Surgery. Conf Proc IEEE Eng Med Biol Soc 2008:1:3273-76 (full length peer-reviewed proceedings manuscript).

[78] M. Mahvash, and P. E. Dupont, Bilateral teleoperation of flexible surgical robots, IEEE International Conference on Robotics and Automation, Workshop on New Vistas and Challenges in Telerobotics 2008

[79] Zenati MA, Wolf A, Ota T, Choset H. High Dexterity Snake Robot for Intrapericardial Interventions. Proceedings, 1st Worldwide Meeting of the Minimally Invasive Robotic Association (MIRA), December 7-10, 2005, Innsbruck, Austria, page 19.

[80] Zenati MA, Ota T, Degani A, Wolf A, Choset H. Percutaneous Subxiphoid Left Atrial Appendage Obliteration with an Innovative Highly Articulated Teleoperated Catheter. Proceedings, I2 Summit 2006 – Innovations in Interventions (co-sponsored by ACC/SCAI), Georgia World Congress Center, Atlanta, GA, March 11-13, 2006, p. 324.

[81] Zenati MA, Ota T, Wolf A, Choset H. Epicardial Atrial Ablation Using a Novel Highly Articulated “Snake” Robot Via a Percutaneous Subxiphoid Approach. Innovations 2006; 1(4):194.

[82] Ota T, Degani A, Wolf A, Choset H, Zenati MA. Epicardial Atrial Ablation Using a Novel Highly Articulated Robotic Probe Through a Subxiphoid Approach. Am J Cardiol 2006; 98(8):248M.

[83] Zenati MA, Ota T, Degani A, Zubiate B, Wolf A, Choset H. An Articulated Telerobotic System for Minimally Invasive Epicardial Interventions. Proceedings, 2nd International Congress Minimally Invasive Robotic Association (MIRA) 2007, New York, NY January 18-20, 2007, page 64

[84] Ota T, Degani A, Zubiate B, Wolf A, Schwartzman D, Choset H, Zenati MA. An Articulated Telerobotic System for Minimally Invasive Epicardial Interventions. Proceedings, 10th Annual NewEra Cardiac Care: Innovation and Technology. Jan 5-7, 2007, Huntington Beach, CA, page 42

[85] Ota T, Degani A, Zubiate B, Schwartzman D, Choset H, Zenati MA. A Highly Articulated Robotic Surgical System for Minimally Invasive Epicardial Interventions. Proceedings, Minimally Invasive Robotic Association, 3rd International Congress MIRA 2008, January 24-26, 2008, Rome, Italy, page 149.

[86] Ota T, Schwartzman D, Nikou C, Jaramaz B, McGarvey J, Zenati MA. Real-time 3D CT Image Guidance and Tool Tracking using an Electromagnetic Tracking System for Minimally Invasive Epicardial Interventions. Innovations 2009;4(3):149.

[87] Yokota T, Ota T, Schwartzman D, Nikou C, Jaramaz B, Zubiate B, Choset H, Zenati MA. Model-Guided Single Port Interventions Using a Novel Highly Articulated Mechatronic System. Proceedings, Annual Meeting of the Biomedical Engineering Society (BMES), Pittsburgh, PA, October 7-10, 2009, page 38.

[88] Yokota T, Ota T, Schwartzman D, Nikou C, Zubiate B, Choset H, Zenati MA. Model-Guided Single Port Epicardial Interventions using a Novel Highly Articulated Mechatronic System. Proceedings, 1st

Worldwide Congress of the Clinical Robotic Surgery Association (CRSA), Chicago, IL October 9-10, 2009, page 19.

[89] Yokota T, Ota T, Zubiate B, Jaramaz B, Zenati MA. Model-guided Single Port Epicardial interventions using a Highly Articulated Robotic System. Proceedings, Minimally Invasive Robotic Association (MIRA) 2010 Meeting, San Diego, CA, January 28-20, 2010. Page 46 (S080).

[90] Choset H, Zenati MA. Enabling Medical Robotics for the Next Generation of Minimally Invasive Procedures. Proceedings, IEEE/ICRA, Anchorage, Alaska, May 2010

[91] Neuzil P, Reddy V, Zenati MA, Kuenzler R, Svanidze O. Single-Port Subxiphoid Access Epicardial Ablation Using “Snake” Robotic System. Innovations 2010;5(3):225 [Proceedings, International Society for Minimally Invasive Cardiac Surgery (ISMICS), 2010 Annual Meeting, Berlin, Germany, June 2010].

[92] C.R. Wagner, N. Stylopoulus, and R.D. Howe, “The Role of Force Feedback in Surgery: Analysis of Blunt Dissection,” Proc. Haptics Symp., Mar. 2002.

[93] D. Pai, “Strands: Interactive simulation of thin solids using cosserat models,” in Proceedings of Eurographics’02, vol. 21, no. 3, 1990, pp.347–352.

[94] S. S. Antman, Nonlinear problems of elasticity. Springer Verlag, New York, 1995.[95] Mahvash, M., Hayward, V. 2004. High Fidelity Haptic Synthesis of Contact With Deformable Bodies.

IEEE Computer Graphics and Applications. 24(2): 48-55, 2004

.