Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Virtual Dosimetrists: A Radiotherapy Training "Flight Simulator"

Skylar S. Gay, Tucker Netherton, Barbara Marquez, Raymond Mumme, Mary Gronberg, Brent Parker, Chelsea Pinnix, Sanjay Shete, Carlos Cardenas, Laurence Court

TL;DR

The paper introduces Virtual Dosimetrist, a cross-modal framework that combines dose distribution prediction with natural language prompts to generate and iteratively refine radiotherapy dose distributions for training purposes. Using 53 head-and-neck VMAT plans, it creates 4072 suboptimal-to-improved dose distributions via a replanning technique, and trains a dual-encoder architecture (3D DDU-Net + CLIP) to map prompts to dose changes in about $7$ seconds per instance. A sliding-window inference with a prompt-refinement loop enables real-time, clinic-like feedback for plan quality review training, independent of a TPS. The approach supports rapid, varied educational examples and has potential to enhance trainee proficiency in plan review and improvement while reducing dependence on traditional clinic workflows.

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.

Virtual Dosimetrists: A Radiotherapy Training "Flight Simulator"

TL;DR

The paper introduces Virtual Dosimetrist, a cross-modal framework that combines dose distribution prediction with natural language prompts to generate and iteratively refine radiotherapy dose distributions for training purposes. Using 53 head-and-neck VMAT plans, it creates 4072 suboptimal-to-improved dose distributions via a replanning technique, and trains a dual-encoder architecture (3D DDU-Net + CLIP) to map prompts to dose changes in about seconds per instance. A sliding-window inference with a prompt-refinement loop enables real-time, clinic-like feedback for plan quality review training, independent of a TPS. The approach supports rapid, varied educational examples and has potential to enhance trainee proficiency in plan review and improvement while reducing dependence on traditional clinic workflows.

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.
Paper Structure (10 sections, 2 equations, 2 figures, 2 tables)

This paper contains 10 sections, 2 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Data Generation, Curation, and Augmentation
  4. Architecture
  5. Inference and Post Processing Techniques
  6. Results
  7. Discussion
  8. Conclusion
  9. Acknowledgments.
  10. Disclosure of Interests.

Figures (2)

  • Figure 1: Overview of model architecture. Unless otherwise specified, all convolution kernels dimension are 3×3×3, with the number of channels indicated in the figure.
  • Figure 2: A generated lower-quality training example (top left) with reduced esophagus sparing. With the prompt "Decrease the mean dose by -900 cGy", a new dose distribution with improved esophagus sparing was then generated (top right) that is similar to the clinical dose distribution (bottom left). The DVH comparison (bottom right) shows how this tool can both generate lower-quality educational examples and subsequently improve the quality of the examples to be similar to the actual clinical plan. Red arrow: region of the esophagus.