See, Plan, Cut: MPC-Based Autonomous Volumetric Robotic Laser Surgery with OCT Guidance

TL;DR

RATS addresses the lack of volumetric intraoperative sensing and calibration in robotic laser surgery by integrating OCT-guided imaging with a calibrated fiber laser and a sampling-based MPC planner operating on OCT voxels. The system combines macro RGB-D and micro OCT with a multistage calibration pipeline to achieve OCT-to-end-effector accuracy of $0.161 \pm 0.031$ mm, a data-driven LTI model with RMSE of $0.231 \pm 0.121$ mm, and closed-loop volumetric resection with RMSE of $0.842$ mm and IoU gains of $64.8\%$ versus feedforward. OCT enables subsurface structure detection and planner objective reweighting to preserve critical anatomy, demonstrated in phantom and ex vivo tests. The work highlights a practical, modular path toward autonomous, constraint-aware laser resection applicable to neurosurgical oncology and other soft-tissue procedures.

Abstract

Robotic laser systems offer the potential for sub-millimeter, non-contact, high-precision tissue resection, yet existing platforms lack volumetric planning and intraoperative feedback. We present RATS (Robot-Assisted Tissue Surgery), an intelligent opto-mechanical, optical coherence tomography (OCT)-guided robotic platform designed for autonomous volumetric soft tissue resection in surgical applications. RATS integrates macro-scale RGB-D imaging, micro-scale OCT, and a fiber-coupled surgical laser, calibrated through a novel multistage alignment pipeline that achieves OCT-to-laser calibration accuracy of 0.161+-0.031mm on tissue phantoms and ex vivo porcine tissue. A super-Gaussian laser-tissue interaction (LTI) model characterizes ablation crater morphology with an average RMSE of 0.231+-0.121mm, outperforming Gaussian baselines. A sampling-based model predictive control (MPC) framework operates directly on OCT voxel data to generate constraint-aware resection trajectories with closed-loop feedback, achieving 0.842mm RMSE and improving intersection-over-union agreement by 64.8% compared to feedforward execution. With OCT, RATS detects subsurface structures and modifies the planner's objective to preserve them, demonstrating clinical feasibility.

Figures (8)

• Figure 1: RATS Platform: Hardware configuration for OCT-guided volumetric laser surgery: The robot-mounted platform integrates a spectral-domain OCT scanner (1310 nm) and fiber-coupled surgical laser (1060 nm), co-axially aligned with OCT through a dichroic mirror. Collimated laser scalpel beam is focused through an achromatic doublet-based focusing lens. An RGB-D camera provides macro-scale information over OCT's micro-scale features. The system is shown targeting a simulated meningioma tumor embedded within vascular tissue, representative of neurosurgical environments requiring sub-millimeter precision and constraint-aware planning biorender_fig1
• Figure 2: System Overview: Experimental setup with a brain phantom for demonstration. The OCT sensor sees the surface and sub-surface structures, and the reconstructed 3D structure is then used to generate a volumetric resection plan. The planner generates a sequence of robot states and laser parameters to cut the tissue with clinically relevant objectives.biorender_fig2
• Figure 3: Laser Calibration Method: Calibration schematic showing OCT frame (blue--green--red), OCT scanning axis, and laser axis alignment. The intersection point defines the laser focal point $\vec{p}_{L}$, estimated from crater centers $\vec{c}_{i}$ at different depths $z_{i}$.
• Figure 4: Laser--tissue interaction (LTI) model estimation.(a) Example of super-Gaussian parameter estimation through fitting a curve to the ground-truth point cloud obtained from OCT B-scans. This process enables estimation of single-point tissue response for a given laser energy. (b) Example of Gaussian and super-Gaussian curve fits on a ground-truth point cloud along the $x$-axis. Gaussian curve estimation does not capture the multimode beam response adequately. (c) Probability density function of residual error from curve fitting ($z_\text{fit}$) with respect to observed ($z_\text{obs}$) point cloud for Gaussian and super-Gaussian fits, at 50% duty cycle with the LTI model ($P = 12.73$).
• Figure 5: System calibration pipeline and coordinate frame hierarchy: The robot end-effector houses the OCT, RGB-D camera, and laser scalpel. Transformations between robot base, end-effector, OCT, camera, and laser focus frames are shown. Includes OCT 2D-to-3D volume registration, OCT-to-end-effector (EE) frame transformation, and laser-to-OCT alignment. Dashed lines indicate calibrated extrinsic transforms ($T^{EE}_{\text{OCT}}$, $T^{\text{OCT}}_{\text{laser}}$, $T^{EE}_{\text{camera}}$). The figure illustrates the reference frame alignment critical for precise imaging and laser actuation.