Fluence Map Prediction with Deep Learning: A Transformer-based Approach

2025-11-10

AI Summary

Conventional inverse optimization in intensity-modulated radiation therapy (IMRT) is computationally intensive and heavily reliant on expert knowledge. Method: This study proposes an end-to-end deep learning framework that directly predicts multi-field 3D dose distributions from input CT images and organ-at-risk segmentations. We introduce a novel 3D Swin-UNETR architecture incorporating hierarchical window-based self-attention to jointly capture fine-grained anatomical details and global spatial dependencies, augmented with self-supervised pretraining to enhance generalization under limited clinical data. Results: Evaluated on clinical datasets, the model achieves a mean R² of 0.95, mean absolute error of 0.035 Gy, and gamma pass rates of 85±10% (3%/3 mm). Dose-volume histogram (DVH) metrics show no statistically significant differences from clinically approved Eclipse plans. The method enables fully automated, high-fidelity dose prediction without iterative optimization, substantially improving planning efficiency and reproducibility.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Computer Vision: Multi-modal Vision
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

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

Abstract

Accurate fluence map prediction is essential in intensity-modulated radiation therapy (IMRT) to maximize tumor coverage while minimizing dose to healthy tissues. Conventional optimization is time-consuming and dependent on planner expertise. This study presents a deep learning framework that accelerates fluence map generation while maintaining clinical quality. An end-to-end 3D Swin-UNETR network was trained to predict nine-beam fluence maps directly from volumetric CT images and anatomical contours using 99 prostate IMRT cases (79 for training and 20 for testing). The transformer-based model employs hierarchical self-attention to capture both local anatomical structures and long-range spatial dependencies. Predicted fluence maps were imported into the Eclipse Treatment Planning System for dose recalculation, and model performance was evaluated using beam-wise fluence correlation, spatial gamma analysis, and dose-volume histogram (DVH) metrics. The proposed model achieved an average R^2 of 0.95 +/- 0.02, MAE of 0.035 +/- 0.008, and gamma passing rate of 85 +/- 10 percent (3 percent / 3 mm) on the test set, with no significant differences observed in DVH parameters between predicted and clinical plans. The Swin-UNETR framework enables fully automated, inverse-free fluence map prediction directly from anatomical inputs, enhancing spatial coherence, accuracy, and efficiency while offering a scalable and consistent solution for automated IMRT plan generation.

Problem

• Accelerating fluence map generation in radiation therapy planning
• Reducing dependence on manual optimization and planner expertise
• Enabling automated IMRT planning directly from anatomical inputs

Innovation

• Transformer-based deep learning for fluence map prediction
• End-to-end 3D Swin-UNETR network using anatomical inputs
• Hierarchical self-attention capturing spatial dependencies automatically