Kinematic Modelling, Trajectory Planning, and Machine Learning for Integrated C-Arm Fluoroscopy and Operating Table Systems

Abstract

This dissertation presents a unified methodological framework that synergistically integrates kinematic analysis, trajectory planning, and machine learning-enabled kinematic solvers to advance modular C-arm fluoroscopy systems coupled with multi-degree-of-freedom (DoF) operating tables. Conventional C-arm devices, though indispensable for intraoperative visualization, remain hampered by constrained maneuverability, operator-dependent variability, and radiation inefficiencies limitations exacerbated in minimally invasive surgery (MIS), where procedural precision and radiation stewardship are cardinal. To redress these deficiencies, the research unfolds in three interlinked contributions. First, rigorous forward and inverse kinematic formulations, augmented by voxelized collision-aware workspace mapping, are deployed across 6-DoF to 9-DoF C-arm with operating table configurations to quantify anatomical accessibility and collision-free pose feasibility across six clinically relevant projections. Second, multiple trajectory planning paradigms including polynomial interpolation, trapezoidal velocity profiles, and biologically inspired minimum-jerk and minimum-snap strategies are comparatively analyzed to ensure smooth, dynamically admissible, and singularity-robust transitions between imaging poses. To operationalize these strategies intraoperatively, lookup tables (LUTs) derived from precomputed trajectories and singularity analyses are introduced as pragmatic decision-support tools, enabling rapid recall of optimized motion plans tailored to patient-specific anatomy. Third, machine learning (ML) frameworks are harnessed to overcome the computational intractability of traditional numerical solvers for high-DoF systems. Leveraging expansive simulation-derived datasets, five supervised models including deep neural networks are trained and validated, achieving sub-millimetric positional accuracy and sub-degree angular precision while delivering real-time inference that surpasses conventional methods in scalability, robustness, and computational latency. Collectively, these contributions delineate a clinically translatable paradigm that unifies kinematic modelling, trajectory optimization, and ML-based inference, establishing the foundation for intelligent, precision-driven, and radiation-conscious intraoperative imaging within next-generation hybrid operating rooms.