An integrated stereotaxic injection architecture for mouse intracranial surgeries with AI-driven domain expertise

Mouse intracranial microinjection plays a crucial role in neuroscience research, relying on precise hardware, operation, and expertise. Manual injections are error-prone, while commercial automated injectors are prohibitively expensive. Moreover, the required knowledge for precise intracranial injections and surgeries is fragmented. In this study, we present the UD Neuroinjector software-hardware-knowledge architecture aimed at reducing costs and enhancing the planning of stereotaxic microinjections in mice.

Our UD Neuroinjector is constructed using affordable hardware and 3D-printed components, costing only $300, a fraction of the price of commercial injectors. The system is controlled using an Arduino-based firmware, allowing for programmable flow rates, automated injection and extraction, and manual joystick control. Through capillary-fluid measurements, we observed stable and highly linear displacement (R2 = 0.93) at typical injection rates.

To compare our system with existing commercial injectors, we performed direct in vivo DAPI injections into the mouse thalamus, which demonstrated equivalent spread and cell counts. The web-based components of our architecture offer AI-driven assistance and atlas-guided navigation for effective experimental planning. To support experimental design, we employed a hybrid retrieval-augmented generation (RAG) approach, leveraging a corpus of 3,738 scientific papers.

By utilizing the Common Coordinate Framework version 3 (CCFv3), we developed an interactive tool for identifying target coordinates within the mouse brain. Overall, our UD Neuroinjector architecture consolidates and facilitates efficient knowledge discovery, experimental design, planning, and implementation of mouse intracranial surgeries.