Systematic Review of Swarm Intelligence for Autonomous Radiotherapy Planning Decision-Making

Radiotherapy has been a cornerstone of cancer treatment for over a century, yet optimizing its application remains a complex clinical challenge due to patient-specific factors and the intricate interplay of clinical variables. This study investigates the potential of integrating autonomous systems—specifically, nature-inspired swarm intelligence and machine learning (ML) approaches—into radiotherapy planning to enhance treatment efficacy and patient safety. Through a systematic review of existing literature, we compared swarm based autonomous systems with ML methods in radiotherapy planning, focusing on decision-making efficiency, robustness, and computational complexity. Our analysis revealed that ML methods offer significantly faster planning times, reducing computation by up to 70% compared to traditional methods, due to their scalability and efficiency (time complexity of O(N × D)). In contrast, swarm-based methods, while exhibiting longer computation times (up to 50% more than ML methods with a time complexity of O(Np × I × D)), demonstrated superior robustness in handling complex, adaptive scenarios such as real-time adjustments to patient movement or tumor changes. These findings suggest that ML methods are preferable for standard cases requiring expedient planning, whereas swarm inspired methods are more suitable for complex cases where adaptability and robustness are critical. This study contributes to the field by delineating specific contexts in which each autonomous system excels, providing a foundation for future research to develop hybrid models that leverage the strengths of both approaches, ultimately aiming to optimize radiotherapy planning and improve patient outcomes.