Magnetic Soft Catheter Robot System for Minimally Invasive Treatments of Articular Cartilage Defects

Abstract

Articular cartilage defects are among the most common orthopedic diseases, which seriously affect patients’ health and daily activities, without prompt treatment. The repair biocarrier-based treatment has shown great promise. Total joint injection and open surgery are two main methods to deliver functional repair biocarriers into the knee joint. However, the exhibited drawbacks of these methods hinder their utility. The repair effect of total joint injection is unstable due to the low targeting rate of the repair biocarriers, whereas open surgery causes serious trauma to patients, thereby prolonging the postoperative healing time. In this study, we develop a magnetic soft catheter robot (MSCR) system to perform precise in situ repair of articular cartilage defects with minimal incision. The MSCR processes a size of millimeters, allowing it to enter the joint cavity through a tiny skin incision to reduce postoperative trauma. Meanwhile, a hybrid control strategy combining neural network and visual servo is applied to sequentially complete the coarse and fine positioning of the MSCR on the cartilage defect sites. After reaching the target, the photosensitive hydrogel is injected and anchored into the defect sites through the MSCR, ultimately completing the in situ cartilage repair. The in vitro and ex vivo experiments were conducted on a 3D printed human femur model and an isolated porcine femur, respectively, to demonstrate the potential of our system for the articular cartilage repair.

Introduction

The repair of articular cartilage defects is one of the most challenging clinical problems in orthopedics.¹ According to statistics, approximately 900,000 individuals suffer from articular cartilage injuries in the United States every year,² causing a substantial burden on the health care system. Articular cartilage defects often lead to debilitating joint pain, dysfunction, and degenerative arthritis.³ Without early surgical intervention, further joint deterioration becomes inevitable and eventually results in the need for total joint replacement. Over the years, numerous efforts have been exerted to develop efficient cartilage repair techniques, but clinical challenges still remain to be overcome.

Marrow stimulation techniques and repair biocarrier-based treatments are currently two common types of cartilage repair methods in clinical practice. Marrow stimulation techniques, such as abrasion arthroplasty, drilling, and microfracture, can promote the migration of mesenchymal stem cells (MSCs) from the bone marrow to the cartilage defect area and differentiate into fibrocartilage tissue.⁴ However, these tissues possess poorer biomechanical properties than hyaline cartilage.^5,6 Hence, the marrow stimulation technique can only be used as a short-term method to delay cartilage degradation.^7,8 Repair biocarrier-based treatment aims to achieve defect restoration by delivering repair biocarriers, such as therapeutic hydrogel, drugs, or cells, into the joint cavity. Total joint injection and open cavity surgery are used to deliver repair biocarriers. Total joint injection enables the delivery of repair biocarriers from the joint space into the joint cavity, with a minimal incision. Nonetheless, due to the poorly targeted injection and lack of anchoring methods, repair biocarriers will randomly disperse in the entire joint cavity in this process, and only a few of them will act on the defect, leading to unstable repair results. Repair biocarriers can be anchored onto the defect through open surgery. However, the huge postoperative trauma to the patients is inevitable and a long recovery time is required. In recent years, a number of researchers have sought to address these challenges and have obtained some achievements. For instance, Go et al.⁹ developed magnetic microrobots loaded with MSCs and demonstrated the ability to deliver MSCs to cartilage defects through a minimally invasive treatment. However, due to the low MSC load rate of individual microrobots and the difficulty of the microrobot swarm control in the articular cavity, anchoring sufficient amount of repair carriers at the defect sites becomes challenging, thereby diminishing the effectiveness of cartilage repair.

Soft catheter robots, compared with microrobots, can establish a pathway from outside the body to the lesion sites, enabling diagnosis, surgery, or drug delivery at the target location. For example, Pittiglio et al.^10,11 developed magnetic catheter robots for pulmonary endoscopy and targeted photothermal cancer therapy in peripheral lungs. Kim et al.^12,13 presented a submillimeter-scale, self-lubricating magnetic soft continuum robot and demonstrated its potential in the treatment of cerebral aneurysms and ischemic stroke. Moreover, a magnetic-controlled soft catheter system for in situ printing on the surface of human organs was developed by Zhou et al.,¹⁴ and its feasibility was validated on the surface of rat livers. In addition, soft catheter robot systems have shown great potential for disease diagnosis and treatment in different areas,^15,16 such as cardiothoracic,^17–19 ocular,^20–22 and colonic,^23–25 to name a few.

In this study, we develop a magnetic soft catheter robot (MSCR) system (MSCRS) for minimally invasive treatments of articular cartilage defects (Fig. 1A). Compared with the current clinical methods of cartilage repair, through the magnetic navigation of a millimeter-scale MSCR and curing of repair hydrogels (Fig. 1B and 1C), we can make the repair hydrogels accurately and efficiently act on the defect with a minimal postoperative trauma. To control the movements of MSCR in the articular cavity, we develop a hybrid control strategy combining neural network and visual servo, which enables the MSCR to accurately reach different target positions within the articular cavity. Once the robot reaches the site of the cartilage defects, we can inject repair biocarriers through the hollow injection channel inside the robot and anchor the biocarriers in place through photopolymerization, ultimately completing the in situ repair task. The feasibility of this method is validated through experiments conducted on a 3D printed human femur model and an isolated porcine femur.

FIG. 1.

An overview of the MSCRS for minimally invasive treatments of knee joint cartilage defects. (A) Knee joint of the patient placed in the workspace of the MSCRS. (B) MSCR guided to the cartilage defect site by the control subsystem. (C) Repair hydrogel injected to the defect site and then anchored using a beam of 405 nm light.

Materials and Methods

Configuration of the MSCRS

As shown in Figure 2A, the MSCRS consists of a millimeter-scale MSCR and a control subsystem. The workspace is within a sphere of 100 mm diameter. Detailed information on the two parts will be given below.

FIG. 2.

Magnetic soft catheter robot system (MSCRS) for articular cartilage repair. (A) An overview of MSCRS. The MSCRS consists of a millimeter-scale MSCR and a control subsystem composed of a magnetic actuating module and an advancer module. (B) Knee model with minimal incision placed in the workspace of MSCRS. The MSCR is inserted into the knee joint through a minimal incision by the cooperative control of the magnetic actuating module and the advancer module. The injection tube and the fiber are integrated into the MSCR to repair the defect. The surgery is visualized by the microcamera at the tip. (C) Advancer module that controls the propulsion distance of the MSCR through a motor. A syringe pump and two light source generators are integrated inside the advancer module to achieve the corresponding functions.

Design of the MSCR

The main body of the MSCR is a flexible silicone hollow tube with an outer diameter of 4 mm, integrating some functional components internally: (1) an annular permanent magnet (N54, Zhejiang Xirui Magnetic Material Co., Ltd., China), (2) a microcamera (OV6949, Shenzhen Yichuang Electronics Co., Ltd., China), (3) an injection tube, and (4) an optical fiber (Zhongshan Aoxiangguang Technology Co., Ltd., China) (Fig. 2B). The annular permanent magnet at the tip of the silicone tube can cause the deflection motion of the MSCR under external magnetic field. The microcamera provides visual feedback for precise targeting of the MSCR to the cartilage defect sites. When the tip of the MSCR is aligned with the defect sites, the repair hydrogel prepolymer can be injected into the defect area through the injection tube. The optical fiber transmits two types of lights for the motion control of the MSCR and the cartilage repair. During the motion control stage of the MSCR in the joint cavity, white light is transmitted in the optical fiber to provide illumination lighting for the visual servo guided by the microcamera. During the cartilage repair stage, 405 nm light is transmitted in the optical fiber for curing and anchoring the repair hydrogel prepolymer onto the defect sites through photopolymerization reaction.

Control subsystem

The control subsystem of the MSCRS comprises a magnetic actuating module and an advancer module, which control the deflection motion and propulsion distance of the MSCR in the joint cavity, respectively.

The magnetic actuating module consists of five static electromagnetic coils. The magnetic torque $T_{m}$ applied on the permanent magnet of the MSCR is determined by the following equation: $T_{m} = m \times B = [\begin{matrix} 0 & - m_{z} & m_{y} \\ m_{z} & 0 & - m_{x} \\ {- m}_{y} & m_{x} & 0 \end{matrix}] [\begin{matrix} B_{x} \\ B_{y} \\ B_{z} \end{matrix}]$ (1)where $m \in R^{3}$ is the spatial magnetization vector of the permanent magnet, and $B \in R^{3}$ is the magnetic field vector generated by the magnetic actuating module. Then, we can acquire the magnetic field at a specified location point as follows: $B (p) = b (p) I$ (2)where $p \in R^{3}$ is the specified location point in the workspace, $b \in R^{3 \times 5}$ is the current normalized field at the specified point, and $I \in R^{5}$ is a vector of currents.

The electromagnetic coils are powered by five programmable direct current power supplies (IT6942B, ITECH Electronic Co., Ltd., China). These power supplies are connected to a personal computer using the commands of standard format for programmable instruments for remote communication.

The advancer module controls the propulsion distance of the MSCR by a motor (Fig. 2C). This module includes two interchangeable light source generators for emitting the white illumination light and the 405 nm light, and a syringe pump for controlling the injection of photosensitive hydrogel prepolymer into the cartilage defect sites.

Through the collaborative work of the magnetic actuating module and the advancer module, 3D motion control of the MSCR can be achieved, allowing it to flexibly reach the cartilage defect sites and perform the repair tasks.

Hybrid control strategy combing neural network and visual servo

After making a small incision on the skin and puncturing the joint capsule with a blade, the MSCR can enter the joint cavity through the incision to perform repair work. However, not all superficial sites can be selected for incision due to the anatomy of the knee joint (Fig. 3A). During minimally invasive surgery, ligaments, meniscus, blood vessels, and healthy cartilage within the joint cavity should be avoided. In clinical practice, most minimally invasive knee joint surgeries use anteromedial and anterolateral portals (Fig. 3B). Therefore, for some hard-to-reach defects, long navigation trajectories and appropriate deflection are needed for the MSCR to reach the defect sites (Fig. 3C). Here, we divide the motion trajectory of the MSCR from the incision to the defect sites into two segments: the first segment refers to the MSCR reaching the vicinity of cartilage defect from the external skin incision; the second segment refers to the MSCR moving from the vicinity of the cartilage defect to the target defect position. In the first segment, the motion distance of the MSCR is relatively long, and high-efficiency movement is required. In the second segment, the motion distance is relatively short, and high positioning accuracy is required. Therefore, we designed a hybrid control strategy combining neural network and visual servo to complete the coarse positioning and the fine positioning for the two segments, respectively. Notably, the two control processes are independent.

FIG. 3.

Hybrid control strategy of the MSCRS. (A) Anatomical structure of knee joint. (B) Anteromedial and anterolateral portals for minimally invasive surgeries. (C) Positioning process of the MSCR consists of two steps: coarse positioning using neural network-based control and fine positioning using visual servo control. (D) Structure of the BP neural network used to learn the inverse mapping of the nonlinear system. (E) Workflow of visual servo control.

Neural network-based control for long-distance coarse positioning

The MSCRS is a high-order nonlinear and time-varying system, and the functional components in the MSCR introduce a number of uncertain parameters, rendering difficulty to obtain the inverse mapping from position space to actuation space through mathematical calculations. Network-based method^26–29 is widely utilized as a powerful tool due to their high precision and low computational costs. Current studies^30,31 demonstrate the feasibility and advantages of this method in controlling magnetic continuum robots. In this study, we report a neural network-based method to train this inverse mapping of the MSCRS. Specifically, the mapping from the position space to the actuation space can be regarded as a regression problem, and the backpropagation neural network is used as an offline learning method here.

As shown in Figure 3D, we take the 3D position coordinates $(x, y, z)$ of the MSCR’s tip as input, and the desired spatial magnetic field $(B_{x}, B_{y}, B_{z})$ of the magnetic actuating module and the propulsion distance $λ_{m}$ of the advancer module as outputs. The error function $E (w)$ and the target vector $f_{i}$ are defined as follows: $E (w) = \frac{1}{2} \sum_{i = 1}^{n} {(y_{i} - f (x_{i}, w))}^{2}$ (3) $f_{i} = y_{i} - f (x_{i}, w)$ (4)where $n$ represents the total number of data samples input to the network, $y_{i}$ represents the true value of the $i$ -th data sample, and $f (x_{i}, w)$ denotes the predicted output value of the network for the $i$ -th sample. The Jacobian matrix $J_{i j}$ is defined as follows: $J_{i j} = \frac{\partial f (x_{i}, w)}{\partial w_{j}}$ (5)where $w_{j}$ represents the $j$ -th weight in the network. The Levenberg–Marquardt (L–M) algorithm is used to update the weight matrix, as follows: $Δ w = - {(J^{T} J + μ I)}^{- 1} J^{T} f$ (6)where $μ$ is the damping factor in the L–M algorithm.

For data collection, 913 location points were selected within a sphere of approximately 40-mm diameter in the center of the workspace, with a distance of 4 mm between the points. We moved the tip of the MSCR to each selected point and recorded the position coordinates $(x, y, z)$ , spatial magnetic field $(B_{x}, B_{y}, B_{z}),$ and the propulsion distance $λ_{m}$ . The position coordinates of the tip were measured by a binocular camera, and the magnetic field can be calculated from the current of each coil according to the Equation (2). Approximately 70% of the location points were used as the training set, 15% as the test set, and 15% as the validation set. The neural network consists of a hidden layer with 10 neurons. With the inverse mapping acquired by the machine learning, we can map the desired trajectory to the required control variables (including magnetic field generated by the magnetic actuating module and the propulsion distance of the advancer module) and actuate the MSCR to reach the vicinity of the defect.

Visual servo control for short-distance fine positioning

When the MSCR moves to the vicinity of the defect, that is, the defects appear in the field of view of the microcamera, we adopt a visual servo control method to precisely move the MSCR to the target defect position defined by the operator in practice. The workflow of the visual servo control is shown in Figure 3E.

We use the strategy of multimodel fusion. Several trackers can be built to track the selected defect features, and the weighted summation strategy was used to fuse the results and determine the final target position. The weight $h_{i}$ of any tracker is defined as follows: $h_{i} = \frac{s_{i}}{\sum_{k = 1}^{n} s_{k}}$ (7) $s_{i} = \{\begin{matrix} n^{- 1}, \\ 0, \end{matrix} \binom{i f P_{i} exist}{else}$ (8)

Then, the coordinates of the target can finally be obtained as $P_{est}$ , as follows: $P_{est} = \sum_{i = 1}^{n} h_{i} P_{i},$ (9)where $n$ is the total number of trackers, and $P_{i}$ indicates the coordinates of the target position obtained by the $i$ -th tracker. Specifically, two trackers based on kernel correlation filter and channel and spatial relatability tracking algorithm have been developed. More trackers can be easily incorporated into the visual servo controller to improve the robustness of tracking using the multimodel fusion strategy. If all the trackers lose the target during treatment, then the MSCRS requires the operator to reselect the defect feature manually.

We hypothesize that the tip of the injection tube is in the center of the image after calibration, and its coordinate is $P_{ref} = (0, 0)$ . Let the coordinate returned by the tracker be $P_{est}$ . We define the control error at each time step $k$ as follows: $e (k) = P_{ref} (k) - P_{est} (k)$ (10)

The proportion—integral—derivative (PID) algorithm is used to realize the position control of the tip. The incremental discrete PID expression is as follows: $u (k) = u (k - 1) + K_{p} [e (k) - e (k - 1)] + K_{i} e (k) + K_{d} [e (k) - 2 e (k - 1) + e (k - 2)]$ (11)where $K_{p}$ , $K_{i}$ , and $K_{d}$ are the proportional, integral, and differential coefficients, respectively. The control quantity $u (k)$ is transformed into the magnetic field vector $B_{g}$ to control the orientation of the MSCR.

Results

Motion characterization of the MSCR under the control subsystem

To investigate the motion characteristics of the MSCR under the control subsystem, we considered the drift angle of the magnetic field ( $γ_{field}$ ) and the inserted length of the MSCR ( $L$ ) as variables and investigate their influence on the offset distance ( $D_{tip}$ ) and deflection angle ( $θ_{tip}$ ) of the MSCR (Fig. 4a). In this study, $γ_{field}$ represented the angle between the magnetic field direction and the Y-axis direction in the system coordinates.

FIG. 4.

Movement of the MSCR under the control subsystem. (A) Sketch of the MSCR under the actuation of the control subsystem. (B) and (C) offset distance ( $D_{tip}$ ) and deflection angle ( $θ_{tip}$ ) of the MSCR under the different drift angles of the magnetic field ( $γ_{field}$ ) and the different insertion lengths of the MSCR ( $L$ ). Each experiment was repeated five times.

We set the intensity of the magnetic field to 8 mT through the magnetic module and controlled the drift angle of the magnetic field ( $γ_{field}$ ), varying from $0^{°}$ to $90^{°}$ . The insertion lengths of MSCR were set to 50, 60, and 70 mm through the advancer module. As shown in Figure 4B and 4C, the $D_{tip}$ and $θ_{tip}$ of the MSCR monotonically increased as a function of the drift angle of the magnetic field $γ_{field}$ for different insertion lengths, $L$ . In addition, longer insertion lengths $L$ resulted in greater offset distance $D_{tip}$ and deflection angle $θ_{tip}$ . The experimental results show that the position and orientation of the MSCR could be smoothly controlled by the magnetic actuating module and advancer module of the control subsystem.

Validation of neural network-based control

Experiments were carried out to validate the effectiveness of the neural network-based control method for the MSCR coarse positioning. The tip of the MSCR was controlled to track a circular trajectory with a diameter of 15 mm (Fig. 5A) and a square trajectory with a side length of 15 mm (Fig. 5C). The motion speeds of the MSCR were set to 0.26 mm/s for circular trajectory and 0.46 mm/s for square trajectory. Figure 5B and 5D show the tracking errors of the two trajectories. The figures show that the proposed neural network-based control method exhibit good performance. The maximum absolute tracking errors for circular and square trajectories are 1.82 and 1.49 mm, respectively, and the mean absolute errors are 0.63 and 0.56 mm.

FIG. 5.

Tracking results of the neural network-based control. (A) and (C) Tracking of circular and square trajectories by neural network-based control, respectively. (B) and (D) Tracking errors for circular and square trajectories, respectively. The red dashed line represents the mean error. Each trajectory was repeated three times.

Validation of visual servo-based control

Experiments were then performed to demonstrate the effectiveness of the visual servo-based control method for the fine positioning of the MSCR. A sheet of article with different feature shapes was placed in the microcamera’s field of view (Fig. 6A). The tip of the injection tube in the camera image was initially aligned with the first star-shaped feature. For ease of description, we temporarily create a coordinate in the article plane and set the star-shape feature as the initial point (0, 0) to describe the change in the tip position. Then, the tip of the injection tube was controlled to align with the remaining target features subsequently, simulating the tracking of different defect sites on the cartilage. Figure 6A and the Supplementary Video S1 show the tracking process of the experiments, and Figure 6B and 6C show the coordinate variation of the tip of injection tube. The proposed visual servo control method exhibits excellent performance with the final steady-state error of less than 0.3 mm.

FIG. 6.

Experimental results of visual servo control. (A) Different features tracked by visual servo control. The red arrows and numbers represent the direction and sequence of the desired movement. The blue coordinate system is a temporary coordinate system to describe the change in the tip position. (B and C) Desired and actual changes in the X and Y coordinates.

Treatment on an in vitro 3D printed human femoral model

To demonstrate the potential application of the MSCRS for the minimally invasive treatments of articular cartilage defects, we conducted confirmatory experiments on a fabricated femoral model (Fig. 7A).

FIG. 7.

Repair experiment performed on a printed femur model with two defects. (A) Experimental setup. (B) Planned trajectories on the printed femoral model. (C) Frames of the two repair processes on the femur model recorded by the external camera and the microcamera integrated into the MSCR at various times.

To reserve the space for the movement of the MSCR and obtain a better presentation, a one-third scale human femur model was fabricated through 3D printing and used for experiment. Two cylindrical defects (Defects 1 and 2 with 12 mm in diameter and 4 mm in depth) were created on the model at different positions to simulate cartilage defects. A piece of silicone-made imitation skin covered a part of the knee model, mimicking the skin tissue outside the joint. A small aperture was created in the imitation skin to simulate the minimal incision in the cartilage repair surgery. The center of the small aperture served as the starting point, and Defects 1 and 2 served as the target points. The femur model with defects was converted into a 3D point cloud dataset to plan the path of the MSCR. The original paths from the starting point to the defects were obtained by the A-star algorithm. The cost function f(n) of the A-star algorithm is described as follows: $f (n) = g (n) + h (n)$ (12)where $g (n)$ is the base cost representing the cost from the starting point to the current planning point, and $h (n)$ is the heuristic cost representing the estimated cost from the current planning point to the target point. The Euclidean distance was chosen to represent $g (n)$ and $h (n)$ , as follows: $g (n) = \sqrt{{(x_{n} - x_{s})}^{2} + {(y_{n} - y_{s})}^{2} + {(z_{n} - z_{s})}^{2}}$ (13)and $h (n) = \sqrt{{(x_{n} - x_{t})}^{2} + {(y_{n} - y_{t})}^{2} + {(z_{n} - z_{t})}^{2}}$ (14)where $(x_{n}, y_{n}, z_{n})$ represents the current planning point, $(x_{s}, y_{s}, z_{s})$ represents the start point, and $(x_{t}, y_{t}, z_{t})$ represents the target point.

Finally, by optimizing the original planned paths with third-order Bessel curves, we obtained two final trajectories (traj 1 and traj 2), which are 28.3 and 32.5 mm long, respectively (Fig. 7B). We set the insertion speed of the MSCR as 1 mm/s in the coarse positioning stage and 0.075 mm/s in the fine positioning stage to obtain higher motion efficiency with sacrificing less motion accuracy.

The repair process was divided into four steps and is shown in Figure 7C and the Supplementary Video S2. In Step 1, the neural network-based controller was exerted to control the MSCR move to the vicinity of the target defect. Then, in Step 2, the visual servo controller was used to actuate the injection point of the MSCR aligning with the defect site. In Step 3, the hydrogel prepolymer (15% w/v) (EFL-GM60, Suzhou Yongqinquan Intelligent Equipment Co., Ltd, China) was injected into the defect sites through the injection tube. Immediately after, in Step 4, the light source was switched to expose the hydrogel prepolymer under the 405 nm light, curing and anchoring the hydrogel prepolymer to the defect sites. Through the experiments, we demonstrated that the MSCR can precisely move to the defect site and complete the repair task. Two tasks of repairing defects were completed in 144 and 141 s, respectively, demonstrating the high efficiency of the MSCRS.

Treatment on an ex vivo porcine femur

An ex vivo experiment on an isolated porcine femur was conducted to further evaluate the feasibility of the MSCRS and is presented in Supplementary Video S3. First, we create a defect (approximately 5 mm in diameter) on the surface of the femoral cartilage by using a scalpel (Fig. 8A). The proposed hybrid control strategy was used to drive the MSCR to a sequentially complete coarse and precise positioning (Fig. 8B and 8C). Then, the repair hydrogel prepolymer (15% w/v, EFL-GM60) was injected and cured (Fig. 8D and 8E). The cured hydrogel can be anchored to the defect site tightly due to the chemical bonding between the photo-generated aldehyde groups and the amino groups on the tissue surface.^32,33 Ultimately, the hydrogel was successfully filled into the defect and well anchored, completing the in situ repair of the defect (Fig. 8F).

FIG. 8.

Cartilage repair on an ex vivo porcine femur. (A) Defect created in the surface cartilage of the femur. (B) and (C) Images showing that MSCR was moved to align with the defect. (D) Hydrogel prepolymer injected into the defect through an injection tube inside the MSCR. (E) Hydrogel anchored to the defect using the 450 nm light. (F) Repaired cartilage defect.

Discussion

Minimal invasiveness and precise biocarrier delivery are the two ideal goals in biocarrier-based cartilage repair methods. In this article, we propose the MSCRS to achieve the two goals simultaneously. The experimental results demonstrate the potential application of our method.

First, to achieve a minimally invasive approach, we designed a flexible MSCR, whose outer diameter is limited to 4 mm. Compared with open surgery in the clinic, we aim to reduce postoperative trauma, healing time, and the probability of complications for patients by reducing the access window of robotic interventions. With these advantages, this minimally invasive treatment based on the MSCRS becomes patient friendly.

Second, to achieve accurate in situ repair of cartilage defects, we designed a control subsystem to achieve remote control of the MSCR. A hybrid positioning control strategy combining neural network and visual servo was developed to achieve the precise navigation of MSCR in joint cavity. This approach can significantly improve the targeting rate of the repair biocarriers compared with the method of total joint injection. Thus, the repair effect is expected to be effectively improved.

Moreover, anchoring the biocarriers to cartilage defects is a challenge in the treatment methods based on biocarriers. Without an effective anchoring method, the repair biocarrier is probably detached from the defect after being delivered on it and dramatically reduces the therapeutic effect. Some chondrocyte-based transplantation surgeries require sealing through periosteal flaps or collagen membranes,^34–36 greatly aggravating the surgical workload and operational difficulty. Some investigators have instructed patients to wear a magnetic anchoring device postoperatively⁹ or to implant additional permanent magnets^37,38 to anchor the magnetic repair biocarriers. Such methods may interfere with the patient’s postoperative daily activities. In this study, we use a beam of 405 nm light to cure the repair hydrogel through chemical bond with the surrounding cartilage surfaces.^32,33 It provides a potential anchoring method for minimally invasive and in situ repair. The inherent porous structure of hydrogel also enables it to be a carrier for other repair materials, such as chemokine, chondroinductive molecule, and cells. Therefore, this anchoring method is more suitable and promising for minimally invasive in situ cartilage repair.

Although we have acquired some achievements with our MSCRS, the current size of the workspace is slightly inadequate for adult articular cartilage repair. In future works, we will improve the system with a larger workspace to meet the requirement of clinical applications. In addition, the network-based control stage is open loop control at present. More technologies can be integrated into the MSCRS, such as real-time in vivo position sensing using FBG (fiber Bragg grating) sensors^10,39–41 or wireless position sensing using magnetic positioning^42–44 to compensate for the lack of feedback during neural network-based control.

Overall, the proposed MSCRS for articular cartilage repair features minimally invasive, precise targeting, and the ability to anchor the repair biocarrier to the defect. The patient’s trauma and recovery time is minimized to the greatest extent, whereas the surgical efficiency is improved in the meantime. Such a repair approach is promising to perform cartilage repair surgery in the form of day surgery, and especially suitable for patients with mild symptoms.

Conclusions

To address the challenges of articular cartilage repair technologies in clinic, we designed a MSCRS to realize minimally invasive and precise in situ articular cartilage repair in this study. This system comprises an MSCR with an outer diameter of 4 mm and a control subsystem. The millimeter scale of MSCR allows it to enter the joint cavity through a tiny skin incision to reduce postoperative trauma. A hybrid control strategy combining neural networks and visual servo was introduced to sequentially complete the coarse and fine positioning of the MSCR. After reaching the target, the repair hydrogel prepolymer is injected into the defect through the MSCR, and then anchored to the defect through the functional components. We eventually validated the effectiveness of this system on an in vitro printed human femoral model and an ex vivo porcine femur to demonstrate the potential of the MSCRS for articular cartilage repair surgery. The intervention of such an MSCRS significantly reduced the surgeon’s involvement during the procedure of surgeries, and enabled safe, stable, and precise in situ cartilage repair.

Future work will further validate the effectiveness and potential of the system by adding other repair materials, e.g., drugs and cells, to the hydrogel and conducting animal experiments.

Footnotes

Authors’ Contributions

J.H. participated in the design and fabrication of the robotic system, experimental procedures, and article writing. Y.H. participated in the design and fabrication of the robotic system, and experimental procedures. S.H. participated in the entire method proposal, and article writing. A.L., W.C., and W.Y. participated in the experimental design. Y.H. and J.F. participated in article writing.

Funding Information

This work was supported by the National Natural Science Foundation of China (Grant No. 61803270), and the Fundamental Research Funds for the Central Universities (Grant No. 226–2023-00087).

Supplementary Material

References

Makris

, Gomoll

, Malizos

, et al. Repair and tissue engineering techniques for articular cartilage. Nat Rev Rheumatol, 2014; 11(1):21–34.

Mithoefer

, McAdams

, Williams

, et al. Clinical efficacy of the microfracture technique for articular cartilage repair in the knee: An evidence-based systematic analysis. Am J Sports Med, 2009; 37(10):2053–2063.

Lories

, Luyten

. The bone-cartilage unit in osteoarthritis. Nat Rev Rheumatol, 2011; 7(1):43–49.

Harris

, Siston

, Pan

, et al. Autologous chondrocyte implantation: A systematic review. J Bone Joint Surg Am, 2010; 92(12):2220–2233.

Welton

, Logterman

, Bartley

, et al. Knee cartilage repair and restoration: Common problems and solutions. Clin Sports Med, 2018; 37(2):307–330.

Cucchiarini

, Madry

. Use of tissue engineering strategies to repair joint tissues in osteoarthritis: Viral gene transfer approaches. Curr Rheumatol Rep, 2014; 16(10):449–449.

Kon

, Gobbi

, Filardo

, et al. Arthroscopic second-generation autologous chondrocyte implantation compared with microfracture for chondral lesions of the knee: Prospective Nonrandomized Study at 5 Years. Am J Sports Med, 2009; 37(1):33–41.

Saris

, Vanlauwe

, Victor

, et al. Treatment of symptomatic cartilage defects of the knee: Characterized chondrocyte implantation results in better clinical outcome at 36 months in a randomized trial compared to microfracture. Am J Sports Med, 2009; 37 Suppl 1:10S–19S.

, Jeong

S-G

, Yoo

, et al. Human adipose-derived mesenchymal stem cell-based medical microrobot system for knee cartilage regeneration in vivo. Sci Robot, 2020; 5(38).

10.

Pittiglio

, Chandler

, da Veiga

, et al. Personalized magnetic tentacles for targeted photothermal cancer therapy in peripheral lungs. Commun Eng, 2023; 2(1).

11.

Pittiglio

, Lloyd

, da Veiga

, et al. Patient-specific magnetic catheters for atraumatic autonomous endoscopy. Soft Robot, 2022; 9(6):1120–1133.

12.

Kim

, Genevriere

, Harker

, et al. Telerobotic neurovascular interventions with magnetic manipulation. Sci Robot, 2022; 7(65):eabg9907.

13.

Kim

, Parada

, Liu

, et al. Ferromagnetic soft continuum robots. Sci Robot, 2019; 4(33).

14.

Zhou

, Yang

, Wang

, et al. Ferromagnetic soft catheter robots for minimally invasive bioprinting. Nat Commun, 2021; 12(1):5072.

15.

Lin

, Chen

, He

, et al. Position and orientation control of multisection magnetic soft microcatheters. IEEE/ASME Trans Mechatron, 2023; 28(2):907–918.

16.

Yang

, Yang

, Cao

, et al. Magnetically actuated continuum medical robots: A Review. Advanced Intelligent Systems, 2023; 5(6).

17.

Chautems

, Lyttle

, Boehler

, et al. Design and evaluation of a steerable magnetic sheath for cardiac ablations. IEEE Robot Autom Lett, 2018; 3(3):2123–2128.

18.

Jeon

, Hoshiar

, Kim

, et al. A magnetically controlled soft microrobot steering a guidewire in a three-dimensional phantom vascular network. Soft Robot, 2019; 6(1):54–68.

19.

Richter

, Venkiteswaran

, Misra

. Concentric tube-inspired magnetic reconfiguration of variable stiffness catheters for needle guidance. IEEE Robot Autom Lett, 2023; 8(10):6555–6562.

20.

Charreyron

, Boehler

, Danun

, et al. A magnetically navigated microcannula for subretinal injections. IEEE Trans Biomed Eng, 2021; 68(1):119–129.

21.

Charreyron

, Gabbi

, Boehler

, et al. A magnetically steered endolaser probe for automated panretinal photocoagulation. IEEE Robot Autom Lett, 2019; 4(2):xvii–xxiii.

22.

, Shen

J-H

, Joos

, et al. Design, calibration and preliminary testing of a robotic telemanipulator for OCT guided retinal surgery. In: 2013 IEEE International Conference on Robotics and Automation. IEEE: Karlsruhe, Germany; 2013; pp. 225-31.

23.

Kim

, You

, Kyung

, et al. Endoscopic surgery robot that facilitates insertion of the curved colon and ensures positional stability against external forces: K-COLON. Int J Med Robot, 2023; 19(3):e2493.

24.

Martin

, Barducci

, Scaglioni

, et al. Robotic autonomy for magnetic endoscope biopsy. IEEE Trans Med Robot Bionics, 2022; 4(3):599–607.

25.

Slawinski

, Taddese

, Musto

, et al. Autonomous retroflexion of a magnetic flexible endoscope. IEEE Robot Autom Lett, 2017; 2(3):1352–1359.

26.

Giorelli

, Renda

, Calisti

, et al. Neural network and jacobian method for solving the inverse statics of a cable-driven soft arm with nonconstant curvature. IEEE Trans Robot, 2015; 31(4):823–834.

27.

, Kang

, Branson

, et al. Model-free control for continuum robots based on an adaptive kalman filter. IEEE/ASME Trans Mechatron, 2018; 23(1):286–297.

28.

, Chen

, Lau

HYK

, et al. Data-Driven methods towards learning the highly nonlinear inverse kinematics of tendon-driven surgical manipulators. Int J Med Robot, 2017; 13(3).

29.

Yip

, Camarillo

. Model-Less hybrid position/force control: A minimalist approach for continuum manipulators in unknown, constrained environments. IEEE Robot Autom Lett, 2016; 1(2):844–851.

30.

Lloyd

, Hoshiar

, da Veiga

, et al. A learnt approach for the design of magnetically actuated shape forming soft tentacle robots. IEEE Robot Autom Lett, 2020; 5(3):3937–3944.

31.

, Sun

, Zhang

, et al. Data‐Driven navigation of ferromagnetic soft continuum robots based on machine learning. Advanced Intelligent Systems, 2023; 5(2).

32.

Hua

, Xia

, Jia

, et al. Ultrafast, tough, and adhesive hydrogel based on hybrid photocrosslinking for articular cartilage repair in water-filled arthroscopy. Sci Adv, 2021; 7(35).

33.

Yuk

, Wu

, Sarrafian

, et al. Rapid and coagulation-independent haemostatic sealing by a paste inspired by barnacle glue. Nat Biomed Eng, 2021; 5(10):1131–1142.

34.

Gomoll

, Probst

, Farr

, et al. Use of a type I/III bilayer collagen membrane decreases reoperation rates for symptomatic hypertrophy after autologous chondrocyte implantation. Am J Sports Med, 2009; 37 Suppl 1(1 suppl):20S–23S.

35.

Gooding

, Bartlett

, Bentley

, et al. A prospective, randomised study comparing two techniques of autologous chondrocyte implantation for osteochondral defects in the knee: Periosteum covered versus type I/III collagen covered. Knee, 2006; 13(3):203–210.

36.

Peterson

, Minas

, Brittberg

, et al. Two-to 9-year outcome after autologous chondrocyte transplantation of the knee. Clin Orthop Relat Res, 2000; 374(374):212–234.

37.

Hori

, Deie

, Kobayashi

, et al. Articular cartilage repair using an intra-articular magnet and synovium-derived cells. J Orthop Res, 2011; 29(4):531–538.

38.

Panseri

, Russo

, Sartori

, et al. Modifying bone scaffold architecture in vivo with permanent magnets to facilitate fixation of magnetic scaffolds. Bone (New York, NY), 2013; 56(2):432–439.

39.

Donder

, Baena

. Kalman-Filter-Based, dynamic 3-D shape reconstruction for steerable needles with fiber bragg gratings in multicore fibers. IEEE Trans Robot, 2022; 38(4):2262–2275.

40.

Pittiglio

, Orekhov

, da Veiga

, et al. Closed loop static control of multi-magnet soft continuum robots. IEEE Robot Autom Lett, 2023; 8(7):3980–3987.

41.

Xiao

, Xu

, et al. Femtosecond laser auto-positioning direct writing of a multicore fiber Bragg grating array for shape sensing. Opt Lett, 2022; 47(4):758–761.

42.

Ferrigno

, Laracca

, Milano

, et al. Magnetic localization system for short-Range positioning: A ready-To-Use design tool. IEEE Trans Instrum Meas, 2021; 70:1–9.

43.

Son

, Gilbert

, Sitti

. Magnetically actuated soft capsule endoscope for fine-Needle biopsy. Soft Robot, 2020; 7(1):10–21.

44.

Song

, Qiu

, Wang

, et al. Design and optimization strategy of sensor array layout for magnetic localization system. IEEE Sensors J, 2017; 17(6):1849–1857.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

1.09 MB

8.23 MB

8.00 MB

0.00 MB