IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO.

12, DECEMBER 2008

2759

A Legged Anchoring Mechanism for Capsule Endoscopes Using Micropatterned Adhesives

Paul Glass, Student Member, IEEE, Eugene Cheung, and Metin Sitti*, Senior Member, IEEE
AbstractThis paper presents a new concept for an anchoring mechanism to enhance existing capsule endoscopes. The mechanism consists of three actuated legs with compliant feet lined with micropillar adhesives to be pressed into the intestine wall to anchor the device at a xed location. These adhesive systems are inspired by gecko and beetle foot hairs. Single-leg and full capsule mathematical models of the forces generated by the legs are analyzed to understand capsule performance. Empirical friction models for the interaction of the adhesives with an intestinal substrate were experimentally determined in vitro using dry and oil-coated elastomer micropillar arrays with 140 m pillar diameter, 105 m spacing between pillars, and an aspect ratio of 1:1 on fresh porcine small intestine specimens. Capsule prototypes were also tested in a simulated intestine environment and compared with predicted peristaltic loads to assess the viability of the proposed design. The experimental results showed that a deployed 10 gr capsule robot can withstand axial peristaltic loads and anchor reliably when actuation forces are greater than 0.27 N using dry micropillars. Required actuation forces may be reduced signicantly by using micropillars coated with a thin silicone oil layer. Index TermsBiologically inspired adhesives, biomedical robotics, capsule endoscopy, gastrointestinal (GI) tract.

I. INTRODUCTION

N ESTIMATED 19 million Americans suffer from diseases of the small intestine such as obscure bleeding, Crohns disease, irritable bowel syndrome, chronic diarrhea, and cancer. However, these pathologies are difcult to diagnose through traditional methods such as wired enteroscopy, wired endoscopy, and radiology [1]. Wireless capsule endoscopes [2][4] have been commercially available since 2001. These pill-sized devices, swallowed by a patient, contain a camera, a battery, and a transmitter and can send high-resolution images of the gastrointestinal (GI) tract to a digital storage device where they can later be analyzed by a clinician. Wireless endoscopes have several advantages over more traditional wired scopes. Most importantly, they allow for direct imaging of the entire small intestine, whereas wired tools can only access the superior and inferior extremities. Unlike capsule endoscopes, wired

Manuscript received November 23, 2007; revised May 12, 2008. First published November 17, 2008; current version published December 17, 2008. This work was supported in part by the Dowd-ICES fellowship program and in part by the Intelligent Microsystem Center. Asterisk indicates corresponding author. P. Glass is with the Biomedical Engineering Department and NanoRobotics Laboratory, Carnegie Mellon University, Pittsburgh, PA 15213 USA (e-mail: pglass@andrew.cmu.edu). E. Cheung is with the Mechanical Engineering Department and NanoRobotics Laboratory, Carnegie Mellon University, Pittsburgh, PA 15213 USA (e-mail: eccheung@cmu.edu). *M. Sitti is with the Mechanical Engineering Department and NanoRobotics Laboratory, Carnegie Mellon University, Pittsburgh, PA 15213 USA (e-mail: sitti@cmu.edu). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TBME.2008.2002111

interventions can also cause discomfort and pain as the scope is manually pushed through the GI tract. While recent studies [5][7] have shown that capsule endoscopy is more accurate than previous diagnostic techniques, they also reveal that some cases of disease are missed by capsule endoscopy due to a lack of control over the position, orientation, and speed of the microcapsule, as it is propelled through the digestive system by peristalsis. Because capsule endoscopes cannot be actively controlled, commercially available devices have no capabilities other than basic imaging or passive monitoring of clinically signicant readings such as pH, temperature, and pressure along the GI tract. If a capsule could be activated and controlled by a clinician to temporarily resist peristalsis and anchor itself to the intestinal lining at a desired location, additional tools for tissue biopsy, drug delivery, and cleaning or cauterizing angiectasias could be incorporated into the device design to enhance its functionality. Recent research has been conducted on incorporating such anchoring mechanisms into microcapsule robots. Kim et al. [8] built and tested a capsule powered by eight individual shape memory alloy (SMA) actuators and embedded with microhooks to provide traction. Later, the same group developed a micromotor-actuated locomotion system for capsules inspired by canoe-paddling [9]. Menciassi et al. [10] and Quirini et al. [11] have proposed several alternative clamping devices to attach a microcapsule to the intestinal wall. In these devices, either a claw-like exure hinge or sets of machined legs are actuated to hold the capsule in place. In this paper, we build upon several previously published [12], [13] versions of this research to present and evaluate the concept of using bio-inspired micropatterned repeatable elastomer adhesives in conjunction with a legged anchoring mechanism for controllable capsule endoscopes. Using these soft elastomer adhesives to hold the capsule in place represents a novel departure from previous designs consisting of sharper metallic clamps and hooks. While there is a wide body of literature on bio-inspired repeatable, micro-/nanopatterned adhesives, only our previous work [14] attempted to optimize these adhesives for creating enhanced friction on wet biological tissues such as the small intestine for medical applications. In this paper, while we build on this previous work, many improvements have been realized. First, micropillar adhesives are tested at a much wider range of preload values to account for the forces induced by capsule deployment. Next, the effect of the thickness of the silicone oil layer on frictional performance is demonstrated. A mathematical model is developed for material performance from experimental data. Finally, in vitro experimental performance of these adhesives is demonstrated on an anchoring capsule prototype inside a fresh porcine small intestine to evaluate its clinical viability.

0018-9294/$25.00 2008 IEEE

Authorized licensed use limited to: Univ of Calif Berkeley. Downloaded on January 21, 2010 at 18:19 from IEEE Xplore. Restrictions apply.

2760

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 12, DECEMBER 2008

While other groups have focused on the development of capsules for locomotion in the GI tract, this paper focuses on a detailed modeling and analysis of the frictional forces required for stable anchoring as a rst step toward this end. Actively controllable locomotion would be important for a clinically useful device, as it would allow a capsule to be precisely positioned within the GI tract. The anchoring mechanism proposed in this study could be directly extended to such a controllable locomotion capsule, as we proposed in [12]. Because of limitations with on-board power and the value of a medical doctors time, the capsule proposed in this paper would be used as a follow-up procedure to a standard passive microcapsule according to the following protocol. First, an existing passive microcapsule is swallowed by a patient as normal and the camera is forced through the GI tract by peristalsis. After the procedure, images are analyzed by a clinician to map areas of interest in the intestine and identify any problematic regions that may require a closer look. Next, the new robotic capsule is swallowed by the patient. Until it is activated, it will be forced by the body through the intestine in the same manner as the passive capsule. Once the capsule reaches the predetermined problematic regions of the intestine, it can be activated remotely by a clinician to anchor at the exact location needed to conduct imaging or any of the supplemental interventions proposed before. Eventually, the intervention could be performed remotely by a doctor over the Internet. As the outline of the paper, the principles and advantages of bio-inspired adhesives and the rest of the proposed design are outlined in Section II. This section continues with the presentation of single-leg and full capsule models of the anchoring mechanism before nishing with a description of fabrication techniques used to construct this prototype. Section III reports the experimental results to evaluate the performance of this design and a single-leg mathematical model. In Section IV, overall capsule performance is examined, while Section V outlines directions for future research in this area. II. ANCHORING MECHANISM DESIGN A. Bio-Inspired Micropatterned Adhesives In this paper, elastomer micropillar adhesives are proposed as a solution for anchoring a capsule in place since they may reduce the risk of injury to the intestinal tissue when compared to sharper clamps and hooks. Recent efforts have been made to understand and develop synthetic patterned micro-/nanobrillar adhesives capable of mimicking the performance demonstrated by certain sticky systems in nature, such as the foot hairs (setae) on the toes of geckos or the bristles on beetle feet [15], [16]. Geckos are able to adhere to surfaces and support their own weight due to the cumulative action of intermolecular forces between each of their millions of setae and the adhering surface [17]. Certain beetles demonstrate even stronger adhesive ability by secreting an oily substance from the bristles on their feet to adhere to surfaces by capillary forces in addition to intermolecular forces [18]. Advantages of these passive adhesion systems are that no additional energy is needed to maintain a bond after adhesion is

Fig. 1. 155 m diameter, 1:1 aspect ratio, and 105 m spaced PDMS micropillars coated with 10 000 cSt silicone oil; optical microscope images of top view (left) and side view (right) are shown.

achieved, and that by using a peeling motion, very little power is needed to detach the adhesive from a surface. They also show promise for biomedical applications since biocompatible materials can be used to form the bers and the soft nature of the microscale hairs will not damage the bodys tissues during adhesion. Because of the rough texture of the intestinal lining due to the presence of villi and microvilli, high-density brillar adhesives seem promising since the contact area between the device and the tissue is signicantly increased over at materials, enhancing friction at the interface to provide more reliable capsule anchoring. While most naturally occurring brillar adhesives have high aspect ratios and are composed of relatively stiff materials to prevent ber clumping when wet, low aspect ratio soft polymer micropillar pancakes demonstrate a similar adhesion enhancement [19]. Because of the relative simplicity in manufacturing these low aspect ratio micropillars out of the biocompatible polymer polydimethylsiloxane (PDMS, Sylgard 184, Dow Corning), this type of adhesive was chosen for this research and is shown in Fig. 1. Viscous and biocompatible silicone oil was also used in some tests to coat the pillars to mimic the highly adhesive abilities of the feet of beetles and crickets. B. Design Principle The microcapsule robot was designed around the use of bioinspired micropillar adhesives to anchor the capsule in place, which requires an actuator to press regions of patterned PDMS into the intestine surface. A simple way to perform this task is with several adhesive-covered legs that can be actuated to stick to the intestine wall and hold the capsule in place during clinical interventions. A computer-aided design (CAD) model of our design is shown in Fig. 2. A hollow cylindrical shell (A) that could contain existing imaging and communication hardware is used to support the anchoring mechanism. Three legs (B) spaced equally around this shell are attached to cylindrical pulleys (C) that are free to rotate. The legs are coated with micropatterned adhesive areas (D) for anchoring. Cables (E) are attached at one end to each upper leg and at the other to the cylindrical shell. A rubber spring (F) connects each leg pulley to the base of the shell. In the capsules initial state, the legs are closed and the cables are slack. When the capsule is actuated (either by an SMA wire or a miniature motor), the cables pull the legs, creating torques

Authorized licensed use limited to: Univ of Calif Berkeley. Downloaded on January 21, 2010 at 18:19 from IEEE Xplore. Restrictions apply.

GLASS et al.: LEGGED ANCHORING MECHANISM FOR CAPSULE ENDOSCOPES USING MICROPATTERNED ADHESIVES

2761

Fig. 2. CAD drawing of the conceptual design of a three-legged anchoring capsule robot with a closeup of a single leg. Labeled components include: capsule shell (A), leg (B), pulley (C), adhesive pad (D), cable (E), and PDMS elastic spring (F).

Fig. 3. (a) Geometric model of a single leg in the unactuated position. (b) Diagram of the two contact modes and the forces acting on the leg. In the angled mode on the left, only the distal tip of the footpad is in contact with the substrate. In the at mode on the right, a continuous portion of the footpad is in contact with the substrate.

on the pulleys causing them to rotate and the capsule legs to open outward. This pulley rotation causes the rubber springs to stretch, inducing counter-torques to oppose the cables. The legs will anchor opening when the torques created by the cables are equal to the torques caused by the stretched springs. When the cables are no longer actuated (and therefore no longer causing resistive torques), the spring torques will cause the legs to return to their original undeployed state. The adhesive pads on the capsule legs will generate a highfriction interface when pushed into the intestine wall. As the feet come into contact with the intestine walls, the legs no longer open but instead apply a preload to the adhesive pads as long as actuation is maintained. The simultaneous deployment of three legs results in stable anchoring. A nal critical design element is the exibility of the capsule legs as they are deployed and come into contact with the intestine wall. The compliance of this leg facilitates both contact and detachment from the intestinal surface. When the leg initially comes into contact with the intestine, it bends and curves to conform to the intestine lining, increasing the surface area of the adhesive region in contact with the tissue. Both normal and shear preloads are present at the adhesiveintestine interface. When the leg closes, this structure provides a peeling motion, signicantly reducing the energy required for detachment. The capsule legs are embedded within the shell wall, so that they will not protrude from the capsule when it is in its closed (passive) state. This will allow the legs to detach from the intestine when the legs are retracted since their embedded state prevents them from involuntarily contacting the tissue. The entire shape is streamlined to prevent any sharp edges from injuring the patient or inadvertently getting caught on the folded intestinal surface. The microcapsule robot must be comfortably swallowed by a patient and must pass through the pyloric sphincter from the stomach to the duodenum, an occasional source of failure seen in passive capsule endoscopes [20]. This hollow shell design can encapsulate existing commercially available imaging equipment, resulting in a device barely larger than current passive capsule endoscopes. Actuator selection will depend on how large a cable force is necessary to gen-

erate sufcient friction forces to resist peristaltic contractions in vivo. C. Modeling Single-leg and full-capsule models were created to help predict the performance and ascertain the viability of this design. The single-leg mathematical model was used to predict the forces between the foot and the intestine for different cable forces. The geometry used in this model is shown in Fig. 3(a). The leg and footpad were treated as a single cantilever beam undergoing large angle deection due to a point load at the tip, an assumption that was veried with force-deection measurements on a leg prototype. The PDMS spring was treated as a linear spring with a spring constant determined by axial beam extension theory. A spring with an elastic modulus of Es , crosssectional area As , and unloaded length Ls has a spring constant of: ks = Es As . Ls (1)

When enough force is applied by the cable on the pulley, the legs will rotate open and push the foot up against a substrate. There are two possible modes of contact with the substrate, as illustrated in Fig. 3(b). In the at mode, a portion of the footpad is in full contact, while in the angled mode, only the distal tip of the footpad touches the substrate. The friction force on the footpad Ffriction is calculated with Ffriction = Fpreload , where is the static friction between the footpad and the substrate. By taking a moment balance on the pulley, the preload force, friction force, leg angle , and contact mode can be calculated for a given cable force. Assuming a uniform pressure on the footpad in the at mode, the preload force was found to be: Fpreload = (r + tl )Fcable rFspring (Ll Lf c x) cos + y sin + 1/2Lf c (r + tf + d) (2)

Authorized licensed use limited to: Univ of Calif Berkeley. Downloaded on January 21, 2010 at 18:19 from IEEE Xplore. Restrictions apply.

2762

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 12, DECEMBER 2008

where r, tf , tl , Ll , and Lf are the mechanism dimensions shown in Fig. 3(a), is the opening angle of the leg, as shown in Fig. 3(b), and x and y are the axial and normal components of the deection of the leg tip relative to its undeformed state. During at mode contact, the length of the footpad in contact is denoted by Lf c , and the force distribution is assumed to be constant across the entire contact area. For angled mode contact, Lf c is equal to 0. For a given cable force, the leg angle and tip deections x and y must be found before the preload and shear forces can be calculated. This is accomplished with an iterative method that steps through possible solutions to these variables and checks that the resulting external forces are consistent with those required to cause the given leg deformation. In at mode, the large angle deection calculations are done with the leg length shortened by Lf c and the tip deections are set to maintain the contact assumption. The previous equations can be adapted to include the effects of a deformable membrane such as intestinal tissue in place of a rigid substrate. A given preload force will cause a deformation xm em in the wall given by a simple spring equation xm em = Fpreload /km em , where km em is a measure of the exural rigidity of the membrane. For a circular membrane of thickness t and radius a pressed at the center km em = 16 a2 Et3 12 (1 2 ) (3)

Fig. 4. Full capsule model in an active intestinal environment. Axial and radial peristaltic loads are designated by P a x and P ra d , respectively, while the adhesion and friction forces imposed by the capsule on the intestinal wall are designated by Fa d and Ffr .

induced foot forces must be able to resist the external peristaltic and gravitational loads to ensure that the capsule remains xed in place for any capsule orientation with respect to a horizontal axis. When the capsule is actuated and motionless within the intestine, static equilibrium results in: Pax + mg sin . (4) 3 In this paper, only the axial peristaltic forces will be considered. This is justied by examining the nature of the radial peristaltic contractions. These oscillations will alternately be acting into and away from the microcapsule body. Because the capsule will be passing through empty, and therefore, collapsed intestines, the radial oscillations acting away from the capsule should not be large enough to separate the adhered feet from the intestine lining. The exibility of the leg design will permit some small oscillations without foot detachment, while radial loads into the capsule will reinforce the adhesion bond. Consequently, the experimental section of this paper will focus exclusively on the forces in the axial direction. Ffr =

where E and are, respectively, the elastic modulus and Poissons ratio of the membrane. Although intestinal material is anisotropic, this isotropic model can be used to make an initial approximation of tissue deformation during preloading. Geometric properties of the tissue are taken from physiological measurements, while mechanical properties are estimated from published values [21]. In our model, this substrate elasticity effectively increases the distance d by the deformation amount xm em . An increase in d will change the preload force applied, so an iterative method is used to determine the equilibrium position. The single-leg model can be incorporated into a full capsule model, as shown in Fig. 4. For deployment in the intestine to be successful, the capsule must be able to resist the peristaltic forces attempting to push the microcapsule robot along the GI tract, while the capsule operator performs a desired stationary task. In numerical simulations of the dynamic electrical and mechanical behavior of intestinal smooth muscle, Miftahof [22] estimates the amplitude of the peristaltic forces to be 17.2 g/cm in the axial direction and 26.9 g/cm in the radial direction. This corresponds to about 460 mN axially and 710 mN radially for a 2.7 cm capsule, which is the length of the most widely used passive capsule endoscope and current prototype. These values will be used as estimates of the loads that a successful anchoring mechanism must be able to resist. Here, these external peristaltic forces acting on the capsule robot are broken up into axial and radial components designated by Pax and Prad , respectively. The adhesion and friction forces induced by the footpad loading the intestine lining for each foot are shown as Fad and Ffr , while the device weight is shown as mg. For this design to be viable, the

D. Fabrication To fabricate the adhesives, an optical lithographic technique is used to pattern microholes into a layer of controlled thickness of SU-8 photoresist that has been spun onto a glass plate. This technique allows for control of pillar density, diameter, and aspect ratio. Preliminary studies [14] indicate that 140 m diameter adhesives with a spacing of 105 m between pillars and a height smaller than 150 m is an optimal geometry for adhesion to the small intestine. A thin layer of PDMS is then molded into these microholes and allowed to cure. Thin strips of Delrin for the capsule legs are lased with a laser engraver (V-Series, Pinnacle) and placed onto the cured polymer layer. Finally, an additional amount of PDMS is poured over the Delrin strips and

Authorized licensed use limited to: Univ of Calif Berkeley. Downloaded on January 21, 2010 at 18:19 from IEEE Xplore. Restrictions apply.

GLASS et al.: LEGGED ANCHORING MECHANISM FOR CAPSULE ENDOSCOPES USING MICROPATTERNED ADHESIVES

2763

Fig. 5. Manufactured capsule prototype with a thin nylon cable (E) providing the actuation force. Letter labels are the same as in Fig. 2.

Fig. 6. Mathematical simulation of the static friction force generated by capsule feet with different coefcients of friction in the small intestine. Friction force increases with increasing cable forces.

cured. Using a sharp blade, the sheets can be cut into individual legs. By maximizing the surface area of the leg covered with the micropatterned material, the resulting friction performance between the leg and the intestine is improved. Dry pillars and wet pillars with tips coated in a 10 000 cSt viscous oil can be fabricated. Tests were conducted with two different oil thicknesses. For the thicker oil layer, a 10 m layer of the oil is spun onto a glass plate. Alternatively, a thin oil lm was created by diluting the oil in a 96% by weight xylene solution before spinning. This signicantly reduces the viscosity of the liquid, allowing it to be spun into a much thinner lm. Xylene evaporates during the spinning process, leaving only the oil lm of original undiluted viscosity. The micropillars are then pressed into the oil layer, transferring a small amount of the liquid onto the pillar tips. The capsule shell is printed out of acrylic in several pieces with a rapid prototyping machine (Invision HR 3-D printer, 3-D systems). Hip joints are similarly printed with cylindrical pulleys to allow the actuation cable to pass over it. Steel pins are inserted through cylindrical slots in both the leg pulleys and the capsule body and are glued into place using Loctite 495 super glue. The PDMS springs for each leg and the multiple shell pieces were similarly glued in place. Actuation cables are passed through the hollow capsule body, around the leg pulley and glued directly to each leg. An assembled capsule prototype is shown in Fig. 5.

A. Model Simulation Results Results of the single-leg model are plotted in Fig. 6. In this plot as well as the experimental results in Section III-B, friction forces are compared with cable actuation forces since this is an important parameter for actuator selection. Because the coefcient of friction between the patterned adhesives and the slippery intestine surface is unknown, a range of values between 0.1 and 0.5 was inserted into the simulation. For all of these modeled materials, the friction forces generated increase approximately linearly with increasing cable forces with proportional increases as the static friction coefcient is raised. Materials capable of generating higher friction forces are desirable, as their use increases the likelihood of successful capsule deployment. In this simulation, the lower friction materials slip past the intestinal surface at specic cable forces, and are no longer in contact with the tissue after that load. For instance, the material with = 0.1 can no longer provide a frictional force for cable loads greater than 0.6 N. According to this model, the friction provided by these footpads alone for the range of cable forces presented here is insufcient to overcome the 460 mN axial peristaltic load required for capsule anchoring. However, this does not take into consideration any macroscale edge effects from the shape of the feet, legs, and capsule body that will also inevitably contribute to the overall capsule friction during deployment. In vitro full capsule testing in a simulated intestine environment will give a better understanding of the overall capsule static friction forces generated. Those tests in conjunction with this model will allow for the identication of the contributions from the footpads as compared with the rest of the capsule. It is important to note that because of the capillary attraction between the moist intestine and the pillars of the adhesives we are using, it is likely that there will exist some friction forces at zero preload, and all of the curves in our model should be offset upward by this initial zero-preload friction value. This offset will also likely increase for micropillars coated with viscous oil, due to the enhanced attraction between the two wetted materials.

III. ANCHORING MECHANISM EVALUATION To evaluate the proposed anchoring mechanism, several experiments were performed and the results of the mathematical model were examined. Friction tests of micropatterned adhesives for increasing preloads were performed to generate empirical equations to compare with our model. Also, in vitro tests of prototype capsule deployment in a simulated simplied intestine environment were conducted.

Authorized licensed use limited to: Univ of Calif Berkeley. Downloaded on January 21, 2010 at 18:19 from IEEE Xplore. Restrictions apply.

2764

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 12, DECEMBER 2008

Fig. 7. Photograph of the single foot friction test apparatus. A PDMS sample is mounted to the underside of the carriage (A) being pulled by a wire (B) mounted to a load cell (C) moving parallel to the plane of the intestine substrate (D). Weights (E) are loaded into the carriage to vary the preload on the foot.

Fig. 8. Friction performance of at PDMS, dry patterned PDMS pillars and oil-coated (both thin lm and thick oil layer) PDMS pillars on freshly harvested porcine intestine. A linear t is superimposed over the data to determine the static friction coefcient for this contact.

B. Friction Characterization of the Micropatterned Adhesive A testing apparatus (see Fig. 7) was constructed to measure the coefcients of friction between the patterned adhesives and a fresh porcine small intestine for increasing leg preloads. A sample of the adhesive material was mounted to the underside of a small carriage (A) that was tied to the stem of a load cell (C, GSO-50, Transducer Techniques) by a thin wire (B). This load cell was mounted to a linear actuator (MFA-CC, Newport Corporation) capable of moving parallel to the intestinal substrate surface (D). Software was written to control the actuator while gathering data from the load cell. All tests were conducted on freshly harvested porcine small intestine samples. These tissue samples were stored in a refrigerated iso-osmotic Tyrodes solution to preserve tissue structure and surface properties before testing. The intestine specimens were slit along their length and clamped as a at sheet to a rigid surface. To maintain tissue hydration and preserve surface properties during testing, the tissue was mounted on a bed of paper towels soaked in Tyrodes solution. The tissue surface was also hydrated with 1 mL of the solution in between tests, with excess uid blotted away. Small weights (E) were placed on the carriage to vary the effective preload on the footpad. For each test, the actuator was retracted at 0.1 mm/s, which is comparable to the speed of propagation of peristaltic waves in the intestine [22]. The subsequent load cell output curves were analyzed to identify the static friction force between the adhesives and the intestinal substrate. This force corresponded to the rst local maxima on the forcetime curve. Because the thin wire pulls directly on the load cell stem without interfering with any other components in the setup, it is assumed that the static friction force between the adhesives and the intestinal substrate was equivalent to the cable force, without losses. Signicant differences were noticed between frictional behaviors when retracting the carriage in the axial versus the transverse intestine directions. Because the capsule will be forced

axially along the GI tract, all data presented in this section were taken from axial friction tests. However, tests in both directions were conducted, and Fig. 7 shows the carriage in the transverse testing conguration. Fig. 8 shows the friction performance of both at and patterned dry PDMS surfaces as well as patterned oil-coated PDMS surfaces for increasing preloads. All of the patterned samples demonstrate enhanced frictional performance with respect to at PDMS. Dry patterned samples showed a 50%100% friction force increase over the at materials over the range of preload forces examined while coating the pillars in oil resulted in a 200%400% friction enhancement over the baseline at samples. For preloads over 0.1 N, pillars coated with a thin lm of oil show higher static friction forces than those coated with a thicker layer. Because the static friction force appears to linearly increase with increasing preload force, these data were used to obtain the static friction coefcient of the contact between the adhesive materials and the intestinal surface for later analysis with the mathematical model. Friction force was linearly related to Fpreload to satisfy the model assumption that Ffriction = Fpreload . No constraint forced the t to pass through the origin because of the possibility for friction to exist in the absence of a preload due to the capillary attraction between the thin liquid layer on the intestine surface and the pillar tips. The obtained linear empirical equations for Ffriction in millinewton are: Ffriction, at = 0.08Fpreload + 13.0 Ffriction, dry patterned = 0.13Fpreload + 27.5 Ffriction, wet patterned = 0.19Fpreload + 106 Ffriction, thin lm patterned = 0.35Fpreload + 89.5 (5)

where the static friction coefcients are obtained from the slope of these ts. Preload forces can be related to cable actuation forces by using (2) to help with eventual actuator selection.

Authorized licensed use limited to: Univ of Calif Berkeley. Downloaded on January 21, 2010 at 18:19 from IEEE Xplore. Restrictions apply.

GLASS et al.: LEGGED ANCHORING MECHANISM FOR CAPSULE ENDOSCOPES USING MICROPATTERNED ADHESIVES

2765

Fig. 9. Emulated intestine environment for measuring the axial loads that the capsule prototype (C) is capable of withstanding when deployed. A length of freshly harvested porcine small intestine (A) is mounted to a tank (B) lled with an iso-osmotic Tyrodes solution. Weights (D) induce cable forces to actuate the legs while the axial load (E) is increased until slipping is observed.

To verify that these adhesives would maintain their frictional performance in uid-lled conditions such as those that may be found in vivo (bile, etc.), this experiment was repeated for dry pillars on a fully submerged intestine, covered with 2 mm deep Tyrodes solution. No signicant change in frictional performance was observed between the two cases. C. Prototype Capsule Deployment To verify our capsule model, a simulated intestine environment (see Fig. 9) was created by mounting a length of freshly harvested porcine intestine (A) to both ends of a uid-lled tank (B). A capsule prototype (C) was inserted into the intestine and deployed with known Fcable values by hanging weights (D) from the leg cables. Through a pulley system, weights (E) were hung from the top of the capsule, creating an axial force between the intestine and the footpads, mimicking the axial peristaltic load present in the GI tract. The magnitude of the axial load was slowly increased until capsule slipping within the mounted intestine was observed. The axial load at this point in time, after subtracting the cumulative axial cable forces, was considered to be equivalent to the axial load the capsule is capable of resisting during deployment. Results from these tests are plotted in Fig. 10. To isolate the contributions of the capsule legs to the overall capsule friction force, the anchoring force of legless capsules was also measured. For dry, patterned footpads, cable forces larger than 0.21 N resulted in total static friction forces over 460 mN, the estimated axial peristaltic load on the capsule. From our full capsule model and (4), this demonstrates the clinical viability of this design for cable forces in this range for capsules oriented at = 0 , while a 10 g capsule oriented at = 90 will require cable actuation forces in excess of 0.27 N for stable anchoring. Because of issues of contaminating the oil layer during deployment, full capsule tests were only performed with dry footpads. Judging from the signicant friction enhancement due to oil-coated pillars as shown in Fig. 8, developing a repeatable method to provide oil to the pillar tips should yield a capsule capable of generating higher static friction forces at comparable cable forces. Using the higher friction coefcient enhancement

Fig. 10. Total capsule static friction force for a deployed capsule in a simulated intestine environment. At actuation forces in excess of 0.21 N, the deployed capsule can withstand axial peristaltic load magnitudes.

result in (5), required cable actuation forces are expected to be much smaller for oil-coated micropillars. Between tests, the patterned capsule legs detached from the collapsed intestinal surface as designed. The peeling behavior of the exible legs, the low adhesive interaction between the legs and the intestine, as well as the embedded nature of the legs in their undeployed position prevented the patterned areas from unintentionally remaining stuck to the intestine lining. IV. DISCUSSION By comparing the predicted capsule forces from the numerical model with the experimental results from the full capsule testing, we can get an estimate of the contributions to the total capsule friction from the different elements of the device design. From our tests, we know that the capsule body alone contributes approximately 120 mN of frictional force at all preloads. From the patterned adhesive friction testing, we know that = 0.13 for the dry patterned adhesives used in our full capsule tests. For cable forces on the order of 0.3 N and = 0.13, the mathematical model provided in (2) can be used to calculate the associated preload force for this adhesive acting on the tissue. The interaction at the footpad can then be calculated from Ffriction = Fpreload to predict that these patterned footpads will contribute about 10 mN to the overall friction force between the capsule and the intestinal lining. Therefore, the remaining 480 mN of anchoring force measured in the deployed capsule experiment is provided by the macroscale edge effects of the capsule legs and feet. As mentioned before, these contributions of the capsule feet are likely underestimated, due to an inability to predict the initial material friction forces at zero preload. Despite this uncertainty in individual friction contributions from various components of the design, the full capsule results appear to demonstrate the clinical viability of these capsules. Material advancements to improve the coefcient of friction

Authorized licensed use limited to: Univ of Calif Berkeley. Downloaded on January 21, 2010 at 18:19 from IEEE Xplore. Restrictions apply.

2766

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 12, DECEMBER 2008

between the patterned adhesives and the intestinal surface will increase the static friction forces generated at specic cable loads and will also allow for viable capsule operation at lower actuation forces. However, it is important to note that the estimated peristaltic forces used in this analysis are based on a numerical simulation, and not on data obtained in vivo. It is possible that the deployment of the capsule legs will induce a more acute peristaltic response than that estimated in [22]. The numerically derived data presented here are the best available approximations of the intestinal force environment, but actual capsule viability cannot be ascertained without in vivo capsule testing. This uncertainty in peristaltic force magnitude is another reason why material advancements to improve the coefcient of friction between the capsule feet and the intestinal lining should be pursued, in order to increase the likelihood that the capsule will remain anchored even while subjected to these uncertain forces. V. CONCLUSION In this paper, mathematical models for a tri-legged anchoring mechanism for enhancing existing passive capsule endoscopes were presented. Through experimentation, these models were validated to help assess the viability of the proposed design with the use of micropillar adhesives. Dry PDMS micropillar samples showed a 50%100% friction enhancement over at materials, while coating the pillars with a thin silicone oil layer resulted in as much as a 400% friction improvement, demonstrating the promise of these bio-inspired materials for microcapsule applications. Full capsule testing showed that this design may be clinically viable with dry adhesives for cable actuation forces in excess of 0.210.27 N, depending on capsule orientation. Comparison of these results with the mathematical model allowed for isolation of the contributions of the pillars themselves from the edge effects of the feet, legs, and capsule body to the overall deployed capsule static friction force. There are several possible avenues for future work in this area including generating more realistic models of the GI tract by incorporating complex tissue anisotropy into the mathematical model, testing recently developed bio-inspired adhesives [23], [24] to determine whether they will yield improved friction between the capsule feet and the intestine lining, and conducting in vivo animal tests to verify the assumptions made in this paper and evaluate the potential of these prototypes for eventual human use. ACKNOWLEDGMENT The authors would like to thank J. Kwon for helping establish the procedure for collecting friction data of materials in contact with intestine substrates, M. E. Karagozler for work on early capsule prototypes actuated by SMA, M. G. Atinc for working with the simulated intestine environment, Dr. R. Appasamy for medical insight, and the remaining staff at the NanoRobotics Laboratory for their continuous feedback and ideas to help advance this research.

REFERENCES
[1] M. Yu, M2A capsule endoscopy. A breakthrough diagnostic tool for small intestine imaging, Gastroenterol. Nursing, vol. 25, no. 1, pp. 24 27, Jan./Feb. 2002. [2] [Online]. Available: http://www.olympus-europe.com/medical/2001_ 5491.htm [3] [Online]. Available: http://www.givenimaging.com [4] [Online]. Available: http://www.rfnorika.com [5] R. Marmo, G. Rotondano, R. Piscopo, M. A. Bianco, A. Siani, O. Catalano, and L. Cipolletta, Capsule endoscopy versus enteroclysis in the detection of small-bowel involvement in Crohns disesease: A prospective trial, Clin. Gastroenterol. Hepatol., vol. 3, no. 8, pp. 772776, Aug. 2005. [6] S. L. Jungles, Video wireless capsule endoscopy: A diagnostic tool for early Crohns disease, Gastroenterol. Nursing, vol. 27, no. 4, pp. 170 175, Jul./Aug. 2004. [7] Z. Fireman and S. Friedman, Diagnostic yield of capsule endoscopy in obscure gastrointestinal bleeding, Digestion, vol. 70, pp. 201206, Dec. 2004. [8] B. Kim, S. Lee, J. H. Park, and J. O. Park, Design and fabrication of a locomotive mechanism for capsule-type endoscopes using shape memory alloys (SMAs), IEEE/ASME Trans. Mechatronics, vol. 10, no. 1, pp. 77 86, Feb. 2005. [9] H. Park, S. Park, E. Yoon, B. Kim, J. Park, and S. Park, Paddling based microrobot for capsule endoscopes, in Proc. IEEE Int. Conf. Robot. Autom., Rome, Apr. 1014, 2007, pp. 33773382. [10] A. Menciassi, A. Moglia, S. Gorini, G. Pernorio, C. Stefanini, and P. Dario, Shape memory alloy clamping devices of a capsule for monitoring tasks in the gastrointestinal tract, J. Micromech. Microeng., vol. 15, no. 11, pp. 20452055, Nov. 2005. [11] M. Quirini, R. J. Webster III, A. Menciassi, and P. Dario, Design of a pill-sized 12-legged endoscopic capsule robot, in Proc. IEEE Int. Conf. Robot. Autom., Rome, Italy, Apr. 1014, 2007, pp. 18561862. [12] M. E. Karagozler, E. Cheung, J. Kwon, and M. Sitti, Miniature endoscopic capsule robot using biomimetic micro-patterned adhesives, in Proc. 1st IEEE/RAS-EMBS Int. Conf. Biomed. Robot. Biomechatronics, Pisa, 2006, pp. 105111. [13] E. Cheung, M. E. Karagozler, S. Park, B. Kim, and M. Sitti, A new endoscopic microcapsule robot using beetle inspired microbrillar adhesives, in Proc. IEEE/ASME Int. Conf. Adv. Intell. Mechatronics, Monterey, 2005, pp. 551557. [14] J. Kwon, E. Cheung, S. Park, and M. Sitti, Friction enhancement via micro-patterned wet elastomer adhesives on small intestinal surfaces, Biomed. Mater., vol. 1, pp. 216220, Oct. 2006. [15] A. K. Geim, S. V. Dubonos, I. V. Grigorieva, K. S. Novoselov, A. A. Zhukov, and S. Y. Shapoval, Microfabricated adhesive mimicking gecko foot-hair, Nature Mater., vol. 2, pp. 461463, Jul. 2003. [16] M. Sitti and R. S. Fearing, Synthetic gecko foot-hair micro/nanostructures as dry adhesives, J. Adhesion Sci. Tech., vol. 17, no. 8, pp. 10551073, May 2003. [17] K. Autumn, M. Sitti, Y. A. Liang, A. M. Peattie, W. R. Hansen, S. Sponberg, T. W. Kenny, R. Fearing, J. N. Israelachvili, and R. J. Full, Evidence for van der Waals attachment for geckos, PNAS, vol. 99, no. 19, pp. 1225212256, Sep. 2002. [18] T. Eisner and D. J. Aneshansley, Defense by foot adhesion in a beetle (Hemisphaerota cyanea), PNAS, vol. 97, no. 12, pp. 65686573, Jun. 2000. [19] A. J. Crosby, M. Hageman, and A. Duncan, Controlling polymer adhesion with pancakes, Langmuir, vol. 21, no. 25, pp. 1173811743, Nov. 2005. [20] E. Rondonotti, J. M. Herrerias, M. Pennazio, A. Caunedo, M. Mascarenhas-Saraiva, and R. de Franchis, Complications, limitations, and failures of capsule endoscopy: A review of 733 cases, Gastrointest. Endosc., vol. 62, no. 4, pp. 712716, Nov. 2005. [21] V. I. Egorov, I. V. Schastlivtsev, E. V. Prut, A. O. Baranov, and R. A. Turusov, Mechanical properties of the human gastrointestinal tract, J. Biomech., vol. 35, pp. 14171425, Oct. 2002. [22] R. N. Miftahof, The wave phenomena in smooth muscle syncytia, In Silico Biol., vol. 5, no. 56, pp. 479498, Nov. 2005. [23] S. Kim and M. Sitti, Biologically inspired polymer microbers with spatulate tips as repeatable brillar adhesives, Appl. Phys. Lett., vol. 89, no. 26, pp. 2691126913, Dec. 2006. [24] B. Aksak, M. Murphy, and M. Sitti, Adhesion of biologically inspired vertical and angled polymer microber arrays, Langmuir, vol. 23, no. 6, pp. 33223332, Feb. 2007.

Authorized licensed use limited to: Univ of Calif Berkeley. Downloaded on January 21, 2010 at 18:19 from IEEE Xplore. Restrictions apply.

GLASS et al.: LEGGED ANCHORING MECHANISM FOR CAPSULE ENDOSCOPES USING MICROPATTERNED ADHESIVES

2767

Paul Glass (S08) received the B.Eng. degree in mechanical engineering (with a minor in arts) from McGill University, Montr al, QC, Canada, in 2005. e He is currently working toward the Ph.D. degree in biomedical engineering at the NanoRobotics Laboratory, Carnegie Mellon University, Pittsburgh, PA. His current research interests include medical robots, biologically inspired adhesives, and minimally invasive surgical technologies. Mr. Glass was awarded a Dowd-ICES Fellowship at Carnegie Mellon University in 2006.

Eugene Cheung received the B.S. degree in mechanical engineering (honors) from California Institute of Technology, Pasadena, in 2003, and the M.S. degree in mechanical engineering in 2005 from the NanoRobotics Laboratory, Carnegie Mellon University, Pittsburgh, PA, where he is currently working toward the Ph.D. degree. His current research interests include biomedical robots and biologically inspired adhesives.

Metin Sitti (S94M00SM08) received the B.Sc. and M.Sc. degrees in electrical and electronics engineering from Bogazici University, Istanbul, Turkey, in 1992 and 1994, respectively, and the Ph.D. degree in electrical engineering from the University of Tokyo, Tokyo, Japan, in 1999. During 19992002, he was a Research Scientist in the Department of Electrical Engineering and Computer Sciences, University of California, Berkeley. He is currently an Associate Professor in the Department of Mechanical Engineering with joint appointments in the Robotics Institute, Electrical and Computer Engineering, and Biomedical Engineering, Carnegie Mellon University, Pittsburgh, PA, where he is also the Director of the NanoRobotics Laboratory. His current research interests include miniature mobile robots, biologically inspired micro-/nanosystems, and micro/nanoscale manipulation and manufacturing systems. Dr. Sitti has been appointed as the Adamson Career Faculty Fellow in 2007. He was the recipient of the National Science Foundation CAREER Award, Carnegie Mellon University Struminger Award in 2005, the second prize in the World RoboCup Nanogram Demonstration League in 2007, the Best Biomimetics Paper Award in the IEEE Robotics and Biomimetics Conference in 2004, the Best Paper Award in the IEEE/RSJ International Conference on Intelligent Robots and Systems in 1998, and the Best Video Award in the IEEE Robotics and Automation Conference in 2002. He was invited as a speaker to the National Academy of Sciences, Keck Foundation Life Engineering Symposium in 2005. He was elected as the Distinguished Lecturer of the IEEE Robotics and Automation Soceity for 20062008. He is the Vice-President of the Technical Activities in the IEEE Nanotechnology Council for 20082010. He is an Associate Editor for the IEEE TRANSACTIONS ON ROBOTICS and the Co-Editor-in-Chief of the Journal of Micro/Nano-Mechatronics.

Authorized licensed use limited to: Univ of Calif Berkeley. Downloaded on January 21, 2010 at 18:19 from IEEE Xplore. Restrictions apply.