Latent Spaces Enable Transformer-Based Dose Prediction in Complex Radiotherapy Plans
This work addresses the resource-intensive process of creating radiotherapy plans for lung cancer patients with multiple lesions, potentially reducing costs and accelerating treatment planning.
The paper tackled the problem of predicting dose distributions for complex multi-lesion lung radiotherapy plans, proposing a two-stage latent transformer framework (LDFormer) that outperforms a state-of-the-art generative adversarial network in dose conformality, especially for overlapping lesions, and generates predictions in under 30 seconds on consumer hardware.
Evidence is accumulating in favour of using stereotactic ablative body radiotherapy (SABR) to treat multiple cancer lesions in the lung. Multi-lesion lung SABR plans are complex and require significant resources to create. In this work, we propose a novel two-stage latent transformer framework (LDFormer) for dose prediction of lung SABR plans with varying numbers of lesions. In the first stage, patient anatomical information and the dose distribution are encoded into a latent space. In the second stage, a transformer learns to predict the dose latent from the anatomical latents. Causal attention is modified to adapt to different numbers of lesions. LDFormer outperforms a state-of-the-art generative adversarial network on dose conformality in and around lesions, and the performance gap widens when considering overlapping lesions. LDFormer generates predictions of 3-D dose distributions in under 30s on consumer hardware, and has the potential to assist physicians with clinical decision making, reduce resource costs, and accelerate treatment planning.