Virtual Dosimetrists: A Radiotherapy Training"Flight Simulator"

2025-05-14

AI Summary

Current radiotherapy plan quality review education is constrained by delayed clinical case updates, insufficient illustrative examples, and limited capability for multi-plan comparative analysis. To address these limitations, this paper proposes a virtual dosimetrist training framework that uniquely integrates deep learning–based dose distribution prediction with natural language understanding (NLU). The framework enables interactive generation and optimization of suboptimal treatment plans: a clinically validated deep learning model synthesizes realistic, substandard dose distributions; a lightweight NLU module interprets natural language instructions; and radiotherapy physics constraints are explicitly embedded to ensure plan feasibility. The system supports millisecond-scale dose recalculation and real-time iterative refinement. Trained on multicenter data, it achieves a threefold improvement in training efficiency. Empirical evaluation demonstrates significant enhancement in trainees’ abilities to identify plan deficiencies, perform multi-objective trade-off analysis, and execute closed-loop optimization—overcoming longstanding bottlenecks in educational resource availability and timeliness.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Machine Learning: Multimodal Learning
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Effective education in radiotherapy plan quality review requires a robust, regularly updated set of examples and the flexibility to demonstrate multiple possible planning approaches and their consequences. However, the current clinic-based paradigm does not support these needs. To address this, we have developed 'Virtual Dosimetrist' models that can both generate training examples of suboptimal treatment plans and then allow trainees to improve the plan quality through simple natural language prompts, as if communicating with a dosimetrist. The dose generation and modification process is accurate, rapid, and requires only modest resources. This work is the first to combine dose distribution prediction with natural language processing; providing a robust pipeline for both generating suboptimal training plans and allowing trainees to practice their critical plan review and improvement skills that addresses the challenges of the current clinic-based paradigm.

Problem

• Lack of robust radiotherapy plan quality training examples
• Inflexibility in demonstrating diverse planning approaches
• Need for accurate, resource-efficient dose modification tools

Innovation

• Virtual Dosimetrist models generate suboptimal treatment plans
• Natural language prompts enable plan quality improvement
• Combines dose prediction with natural language processing