AutoPathNet: a patient-specific automated 3D trajectory planning framework for minimally invasive evacuation of hypertensive intracerebral hemorrhage.

PURPOSE: Rapid and reproducible trajectory planning is important for minimally invasive evacuation of hypertensive intracerebral hemorrhage (ICH). We developed AutoPathNet, a patient-specific automated 3D trajectory-planning framework that generates geometrically favorable candidate trajectories from preoperative imaging. METHODS: AutoPathNet reconstructs patient-specific anatomical models from CT-derived inputs, generates candidate entry-target trajectories, filters candidates according to predefined anatomical constraints, and ranks feasible trajectories using a composite geometric planning metric. Algorithm-generated trajectories were compared with surgeon-implemented catheter trajectories reconstructed from postoperative CT in a 21-patient validation cohort. RESULTS: The framework generated three candidate trajectories within approximately 31 s. In 16 of 21 paired cases (76.2%; Wilson 95% confidence interval, 54.9%-89.4%), the best algorithm-generated candidate had a lower m value than the surgeon-implemented catheter trajectory reconstructed from postoperative CT. This comparison uses postoperative catheter position as a procedural imaging reference and reflects geometric planning performance within the current constraint model, suggesting potential value as a neurosurgeon-supervised planning reference rather than evidence of clinical outcome superiority. CONCLUSION: AutoPathNet rapidly generated patient-specific candidate trajectories for minimally invasive evacuation of hypertensive ICH. The framework may support neurosurgeon-supervised trajectory planning, but prospective validation using multimodal functional constraints and clinical outcome endpoints is required before clinical deployment.