A Rule-Based Multileaf Collimator Algorithm for Dynamic Target Radiotherapy: An Optimized Dmlc Leaf Sequencing Approach
Abstract
Purpose
Accurate and efficient dose delivery is crucial in dynamic multileaf collimator (MLC) radiotherapy to minimize treatment time and reduce delivery errors. Current leaf-sequencing algorithms often face challenges in balancing delivery speed with dose accuracy, especially under complex dose distributions and dynamic target motion. This study aims to develop and validate a novel rule-based MLC leaf-sequencing algorithm designed for dynamic target irradiation. The method incorporates physical leaf constraints to ensure clinically feasible motions while optimizing both total delivery time and dose delivery accuracy.
Methods
The algorithm was evaluated on a simulated dataset (200 samples) and a clinical dataset (24 samples). The simulated dataset was constructed by pairing tumor motion trajectories extracted from 2D cine-MRIs in the TrackRAD2025 dataset with synthetically generated dose distributions. The clinical dataset comprised real tumor trajectories and corresponding dose distributions acquired from 4 patient treatment plans. For each plan, multiple gantry angles were included, yielding a total of 24 distinct sample instances across all plans and beam angles. All data were preprocessed according to standard physical MLC leaf dimensions and machine-specific delivery parameters. Performance was measured using target coverage rate and overdose rate.
Results
The proposed algorithm demonstrated robust performance across both simulated and clinical datasets, consistently achieving high target coverage (>96%) with overdose rates below 11.3% across all tested scenarios. The method also delivered real-time computational performance, with mean processing time under 3 ms per control point.
Conclusion
The proposed Rule-Based MLC Algorithm provides a reliable and efficient solution for dynamic target irradiation in radiotherapy. By integrating row-wise heuristic rules and propagating constraints from a dynamically selected reference row, it effectively handles tumors with diverse dose distributions and motion patterns while adhering to physical leaf constraints. This approach ensures robust performance without compromising delivery accuracy or violating the mechanical limitations of the MLC system.