Autonomous Robotic Drilling System for Mice Cranial Window Creation

TL;DR

The paper tackles autonomous cranial window creation in mice by developing a multimodal robotic drilling system that combines image-based perception with force-driven predictions to handle skull variability. A trajectory planner using constrained splines and online plane fitting enables real-time, overshoot-robust drilling along a circular path, with execution-time updates from a drilling completion level vector c. The authors introduce a deep image-based network and an LSTM-RNN force model to estimate drilling completion, fusing cues to produce high-resolution, occlusion-tolerant planning data, and demonstrate substantial gains in eggshell drilling speed and safety, plus first postmortem mouse skull drilling results. These contributions advance autonomous surgical robotics by providing a generalizable framework that operates without pre-operative data or specialized sensors, and have potential impact for rapid, repeatable cranial window creation in neuroscience research.

Abstract

Robotic assistance for experimental manipulation in the life sciences is expected to enable favorable outcomes, regardless of the skill of the scientist. Experimental specimens in the life sciences are subject to individual variability and hence require intricate algorithms for successful autonomous robotic control. As a use case, we are studying the cranial window creation in mice. This operation requires the removal of an 8-mm circular patch of the skull, which is approximately 300 um thick, but the shape and thickness of the mouse skull significantly varies depending on the strain of the mouse, sex, and age. In this work, we develop an autonomous robotic drilling system with no offline planning, consisting of a trajectory planner with execution-time feedback with drilling completion level recognition based on image and force information. In the experiments, we first evaluate the image-and-force-based drilling completion level recognition by comparing it with other state-of-the-art deep learning image processing methods and conduct an ablation study in eggshell drilling to evaluate the impact of each module on system performance. Finally, the system performance is further evaluated in postmortem mice, achieving a success rate of 70% (14/20 trials) with an average drilling time of 9.3 min.

Figures (11)

• Figure 1: The system setup used in this work consists of one of the 8-degrees-of-freedom robotic branches of an AI-robot platform for scientific exploration Marinho2024. For this work, we attached a micro drill as an end effector. In addition, we use a 4K camera and a clamping mechanism for the object being drilled.
• Figure 2: The clamping mechanism for mouse skulls including a commercial mouse skull holder and a force sensor.
• Figure 3: The clamping mechanism for eggs including a 3D printed holder and a force sensor.
• Figure 4: Block diagram of the proposed autonomous robotic drilling system.
• Figure 5: An example of creating the training dataset. (a) An original image of mouse skull drilling. (b) RGB mask that is labeled manually in 6 classes: Drill, $0\,\%$ drilled (Background), $25\,\mathrm{\%}$ drilled, $50\,\mathrm{\%}$ drilled, $75\,\mathrm{\%}$ drilled, and $100\,\mathrm{\%}$ drilled. (c) Confidence map of the drill. (d) Confidence map of the completion level map. (e) The completion level map is smoothed by a Gaussian filter. (f) Cropped drill mask within the drilling region. (g) Cropped completion level map within the drilling region.