Minimally invasive surgeries (MIS) are gaining importance as they involve smaller incisions, less tissue damage, reduced post-operative pain, and faster patient recovery. Current medical robots — while capable of performing MIS — pose significant disadvantages: limited dexterity, high training requirements, maintenance complexity, size constraints, lack of haptic feedback, and reliability issues.
Our team developed a Computer Vision-Enabled Continuum Medical Robot to perform minimally invasive ureteroscopy with reduced errors and inefficiencies. Continuum robots mimic natural structures, offering tentacle-like flexibility for precise surgical access in anatomical spaces that rigid instruments cannot reach. My focus was the mechanical design, structural validation, actuation integration, and miniaturized hardware packaging.
Three prototype generations were evaluated before arriving at the final architecture — each addressing the prior iteration's fundamental limitations.
| Prototype | Approach | Strengths | Limitations |
|---|---|---|---|
| P1 | Cable-actuated with dual mechanism | Proven actuation concept | Bulky · Complex fabrication · No novelty |
| P2 | Section-view dual mechanism | Improved access geometry | Still bulky · Each mechanism requires separate process |
| P3 (Final) | Unified <12mm continuum platform | Compact · Novel architecture · Simplified assembly | Tight tolerance demands |
Three prototype generations were developed and evaluated before arriving at the final compact architecture. Each iteration addressed specific mechanical and fabrication limitations of the prior design.
Beyond the mechanical platform, the project integrated a full software and sensing stack for enhanced surgical precision and training capability.
- VR-based control system — developed software to control robotic endoscopy via Meta Quest 3 joystick inputs, with real-time video from an onboard camera module streamed into the VR environment via Raspberry Pi and Unity
- Computer vision integration — real-time tracking of phantom kidney position and orientation using ArUco markers for precise AR overlays; mobile imaging captures live surgical feed for navigation guidance
- Phantom testing model — kidney model printed on Stratasys F170 with TPU A 80 material to mimic tissue properties; transparent balloons simulate tissue compliance, adding realism to training scenarios
- Digital twin — visualized robot in 3D Slicer and simulated in Gazebo environment using ROS2; used Open3D and trimesh for high-fidelity rendering of internal anatomical structures
- Custom haptic feedback — attempted to measure end-effector forces during surgery on target organ using strain gauges; developed custom gauges from Polyamide paper, cured silicone sheets, and laser printing due to range limitations of commercial gauges