Shaoping Huang, Zhao He, Yi Chen, Jiafan Chen, Shijie Hong, Yi Zhang, Longyu Xu, Yixin Pan, Shuo Ma, Lian Xuan, Qingdang Meng, Yong Yang, Yangyang Xu, Zecai Lin, Chuqian Lou, Cheng Zhou, Weidong Chen, Bomin Sun, Qingfang Sun, Yuan Feng, Anzhu Gao, Guang-Zhong Yang

Science Advances 2025, 11, eady3624.

Abstract

Stereotactic neurointervention is a common procedure for biopsy, injection, ablation, and implantation of electrodes for deep brain stimulation. Guided by preoperative imaging, conventional approaches are mostly performed manually, lacking operation stability and interactive feedback. The intraoperative magnetic resonance imaging (MRI) guidance enables both structural and functional assessment during operation, permitting interactive adaptation to tissue deformation and avoidance of critical anatomical regions. Here, we report an MRI-guided robotic system for stereotactic neurointervention. A macro-micro hybrid pneumatic-hydraulic actuated stereotactic robot with a large range of motion and high precision is developed. This is coupled with a compact bioinspired soft actuator for target intervention. A global-focal MRI sequence is proposed for interactive navigation, closed-loop control, and precise targeting. Validation is performed with phantom, cadaveric, and in vivo animal studies, showing positional accuracies of 0.39, 0.68, and 0.14 millimeters, respectively, demonstrating superior performance compared to the current state of the art in robotic-assisted stereotactic neurointervention.

Introduction and Methods

Stereotactic neuro-intervention is widely used for procedures such as biopsy, resection, injection, ablation, stereoelectroencephalography, and implantation of electrodes for Deep Brain Stimulation (DBS). Magnetic Resonance Imaging (MRI) is an effective tool for preoperative planning due to its versatility in providing high-contrast soft-tissue characterization and functional information, including perfusion, diffusion, and accurate temperature mapping without ionizing radiation. Intraoperatively, traditional stereotactic neuro-intervention is challenging to perform under MRI guidance due to restricted operational space and MR incompatible hardware that may cause signal loss, field distortion and degradation of SNR. To address these challenges, several MR safe or conditional robotic systems have been developed for stereotactic neuro-intervention.Despite theseadvances, no robotic stereotactic neurointervention systems canthus far achieve fully-actuated stereotactic positioning and performclosed-loop intervention under interactive MRI with submillimeteraccuracy.Current challenges that hinder the accuracy of robotic neuro-intervention systems primarily stem from the following factors. First, the required DoF, large workspace, and deep insertion of the instrument tip impose inherently contradicting constraints for high precision robot control. Second, interventional devices often necessitate long-distance insertion and rotational manipulation for procedures such as biopsy and directional ablation therapy, developing a compact needle insertion mechanism with simultaneous rotation and translation for high-precision target-intervention in a confined space is challenging. Third, balancing temporal and spatial resolution is a major difficulty for intraoperative MRI feedback. In addition, the strong magnetic fields and restrictions of the MR environment, combined with the high sensitivity of MR signals, limit the materials, actuation methods, and sensing approaches that can be used.

In this study, we propose a macro-micro stereotactic robot with high precision and a large range of motion, integrated with a compact bio-inspired soft actuator that utilizes a global-focal MRI sequence for interactive navigation and precise target intervention. A hybrid pneumatic-hydraulic actuation stereotactic system is designed, which includes 4-DoF fast pneumatic macro actuation for global positioning and 4-DoF hydraulic micro actuation for precise local adjustment. A bio-inspired soft actuator is fabricated for target-intervention, with features including compactness (Φ18.4 mm, 30 g), easy to manufacture (3D printing), stable (locking force 13 N) and cost-effective (~$5). Its biomimetic peristaltic motion allows for precise and stable needle insertion. The global-focal MRI sequence allows for interactive navigation and closed-loop control during intervention (global MRI) and precise identification of the approaching target (focal MRI). Detailed functional experiments have been carried out to validate the performance of the system with phantom, cadaveric, and in vivo animal studies.

Key Results and Conclusions

We proposed herein a robotic system for automated stereotactic positioning and closed-loop intervention with submillimeter accuracies. 3D-global MRI was used for stereotactic positioning, while 2D-global and high-resolution focal MRI-guided needle insertion. In phantom experiments, the system achieved a targeting accuracy of 0.39 ± 0.12 mm (N=7), while cadaveric experiments targeting STN showed 0.68 ± 0.13 mm (N=4) accuracy. One in vivo trial showed 0.14 mm targeting error. The hybrid pneumatic-hydraulic actuation system exhibits excellent MR compatibility. The bio-inspired soft actuator for needle insertion resolves key limitations of structural complexity, bulkiness, and poor MR compatibility in decoupled automatic control mechanisms for needle rotation and insertion. The proposed global-focal MRI framework provides an effective strategy for balancing temporal resolution and spatial precision in intraoperative MRI.Compared to existing commercial manual intraoperative MRI-guided stereotactic frames, our system enables automated, stable positioning and insertion. Through MRI-based closed-loop feedback, the system provides interactive visualization during intervention, greatly enhancing the safety and precision of the procedures. The system integrates global-focal MRI for navigation and closed-loop control during intervention and precise identification of the approaching target. It also addresses the limitations of commercial stereotactic frames in clinical applications, which require patients to repeatedly move in and out of the MRI scanner to obtain intraoperative adjustment. Furthermore, compared to existing stereotactic robotic systems, this study achieves for the first time submillimeter positioning accuracy and realizes fully automated large-range positioning and insertion.

Fig.1：Illustration of the MRI-guided robotic stereotactic neurointervention system and its workflow.

(A) The proposed MRI-guided robotic stereotactic neurointervention system consists of a macro-micro robot with hybrid pneumatic-hydraulic actuation and a bioinspired soft actuator guided by global-focal MRI. (B) Overall workflow of the robotic stereotactic neurointervention used in this study. In the preoperative stage, only one MRI scan is required, followed by registration between the MRIand the robot for preoperative planning. During operation, the robot can automatically move along the planned trajectory with high efficiency (fast pneumatic actuationfor macropositioning) and accuracy (accurate hydraulic actuation for fine adjustment) guided by 3D MRI. A burr hole is drilled on the basis of the position on the cranialbone indicated by the robot, and a second anatomical MRI scan is performed to verify the error and tissue deformation caused by “brain shift,” such that further fine adjustments can be made. Once completed, the needle mounted on the bioinspired soft actuator is inserted while the MRI switches from the global mode to local mode foronline monitoring of the instrument trajectory. The path and distance between the needle tip and the target are continuously calculated.

Fig. 2：Design and performance evaluation of the macro-micro hybrid pneumatic-hydraulic actuated stereotactic robot.

(A) Hybrid pneumatic-hydraulic actuatedstereotactic robot. (a), (b), and (c) are used for pneumatic macroactuation, whereas (d) is responsible for linear translation (first df, ±50 mm) and lifting motion (second df,−65° to 20°) through a lead screw. (a) and (b) are conformally integrated on the front and back of the arc, respectively. (a) provides a large arc motion (third df, ±60°),whereas (b) enables rotation on the arc (fourth df, ±20°). (c) provides the microhydraulic actuation which is mounted on the macroactuation between (a) and (b). (B) Detailed MR compatibility tests (SNRloss and distortion) were conducted, and comparisons were made between the control group and different robot operation stages: Staticindicates the robot positioned adjacent to the phantom without power supply; powered represents the powered robot maintaining stationary position; and moving represents the robot executing programmed movements. (C) The fine adjustment results of four microhydraulic joints (x1, y1, x2, y2) based on 16 stereotactic experiments.(D) Histogram of tip error distribution for 39 repeatability tests with a 100-mm insertion length. (E) Force-displacement curves of the robot tip under 5- to 30-­N loads(spaced by 5 N), with each test repeated four times and analyzed using polynomial fitting to evaluate the inherent stiffness of the robot through linear approximation.

Fig.3：Design and performance characterization of the bio-inspired soft actuator.

(A) Illustration of the concept of peristaltic motion driven by travelling waves. (B) The structure of the bio-inspired soft actuator with 2-DoF (translational motion and rotation). (C) A schematic of one chamber of the soft actuator. The helix has a diameter of R+r, where φ indicates the helical angle, λ indicates the length of one period of the chamber, n1-n4 indicate the position of the secondary helixes. (D) The left shows the two pairs of pneumatically driven chambers. The right illustrates the pressure variations in the four outer (CW) chambers. Each chamber is driven by a sinusoidal curve, with phase differences of 90°. (E) Four types of motion based on different travelling wave sequences: forward translation, backward translation, CW rotation, CCW rotation. (F), (G) The FEA strain result for states 2 and 4 of (E), showing four time steps forming one complete cycle. (H) The decoupled translational motion test results. Forward_Rand Forward_T indicate the rotation and the translation during the forward motion test, respectively. Backward_R and Backward _T indicate the rotation and the translation during the backward motion test, respectively. (I) The decoupled rotation test results. CW_R and CW_T indicate the rotation and the translation during the CW rotation test, respectively. CW_R and CW_T indicate the rotation and the translation during the CCW rotation test, respectively. (J) The results of decoupled translational and rotational speed, and the corresponding insertion force during translation in different chamber frequencies (0.5 Hz to 3.3 Hz). (K) The locking force of the needle with helical locking. The graph shows the mean ± s.d. of N = 4 measurements at each data point.

Fig.4： Interactive global-focal MRI.

(A) Global MRI-guided continuous intervention with the robot. The in-plane resolution of global MRI is 1 mm2. (B) Focal MRI-guided step-by-step intervention with the robot. The in-plane resolution of focal MRI is 0.39 mm2. (C) Quantitative comparison of spatial-temporal resolution trade-offs between 2D global and focal MRI. (D) A comparison of full-FOV FSE and rFOV FSE with a fan-shaped phantom. (E) A comparison of full-FOV FSE and rFOV FSE with an in-vivo human brain imaging. (F) Needle tracking results of a simulated brain intervention. (G) The tip and angular errors of needle tracking for the simulated human brain intervention. (H) Needle tracking for an ex-vivo animal tissue intervention. (I) The tip and angular errors of needle tracking for an ex-vivo animal tissue intervention. The video of (F) and (H) can be found in Movie S5.

Fig.5：Phantom validation and cadaveric experiments of the MRI-guided robotic system.

(A) The experiment setup of phantom validation using a 3D-printed skull filled with gelatin containing ten fiducial markers, mounted onto the robot via stereotactic head frame. (B) 2D global MRI during the robotic intervention in a phantom experiment with an in-plane resolution of 1 mm2 and a temporal resolution of 2.5 s/frame. (C) Quantitative analysis of targeting accuracy across seven phantom trials. (D) The experimental setup of cadaver study.(E)Preoperative 3D T2-weighted MRI (0.8×0.8×0.8 mm³) for trajectory planning. (F) Intraoperative 3D T2-weighted MRI (0.8×0.8×0.8 mm³) after craniotomy. (G) MRI-guided needle alignment. (H)Global MRI during robotic intervention for cadaver study. (I)Focal MRI during the robotic intervention. (J) The process of needle alignment in the cadaveric experiment. The robotic system aligned the needle guide with the planned trajectory through task-hierarchical control (Step-2 and Step-3). After the burr hole was created, owing to structural shift, the target and trajectory were updated based on intraoperative MRI. The needle guide was then realigned to the updated path. (K) Continuous distance monitoring between needle tip and target during robotic advancement in the cadaveric experiment. (L) Quantitative analysis of targeting accuracy across four cadaveric experiments.

https://www.science.org/doi/10.1126/sciadv.ady3624