Visuomotor policy enables precise automated microelectrode hooking under a microscope
MicroHookACT: Monocular Microscopic Vision Guided Visuomotor Policy for Flexible Microelectrode Hooking
Robotics
Summary
Placing tiny flexible microelectrodes inside the body requires a tricky step called needle-loop hooking. The authors developed MicroHookACT, a computer program that learns from human demonstrations to control a microscopic camera and robotic actions for this task. It uses visual clues like blur and focus changes to align and hook the microelectrode precisely without touching the surface. Their approach succeeds 96.7% of the time across various setups and completes the task quickly. This shows that teaching robots to see and act like humans can improve precise control in tiny surgical tasks.
What this means in practice
- •For medical device manufacturers: Equip robotic systems for automated microelectrode implantation using monocular microscopic vision and learned visuomotor control.$Commercial implications: Enables product development of automated surgical robots for neuro or bioelectronic device implantation with high accuracy and efficiency.
- •For robotic surgery teams: Implement learned visuomotor policies for precise micron-level robotic actions under simple monocular microscope setups during delicate surgical procedures.
Authors
Yitong Chen, Fangbo Qin, Yang Wang, Ruihua Hu, Kui Zhang, Shan Yu
Abstract
Automated needle-loop hooking is a critical step in flexible microelectrode (FME) implantation. This paper presents MicroHookACT, an imitation learning-based visuomotor policy for automated 3D hooking under monocular microscopic vision. First, a unidirectional hooking strategy exploits defocus cues and optical-axis guidance to enable palpation-free precise alignment and contact-rich threading. Second, an action-supervised object attention module built on a frozen ViT backbone learns to focus on the micro-needle tip and micro-loop directly from human demonstrations, without requiring manual visual annotations for training. Third, attention-centered global coarse and local fine features are dynamically weighted according to predicted action progress, enabling a single ACT policy to adapt to changing defocus blur and visual requirements throughout the operation. In the experiments, visuomotor policies were trained on 60 human demonstrations and evaluated under five setups with varying difficulties. Our MicroHookACT framework achieved the highest overall success rate of 96.7\% with an average execution time of 11.5 s. These results demonstrate the potential of visuomotor policy learning for micron-level control under varying operating conditions.