Dose-Volume Histogram Prediction for Faster Radiotherapy Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern radiotherapy treatment planning faces challenges in achieving optimal dose distribution that balances tumor coverage with minimizing dose to nearby organs at risk (OARs), requiring labor-intensive manual adjustments and tradeoffs between different OARs.
Innovation Solution
A machine-learning-based system predicts dose-volume histograms (DVHs) using overlap-volume histograms (OVHs) and projection-masked OVHs to automate treatment planning, optimizing beam properties and reducing manual goal-setting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If manual adjustment of radiation beam properties is performed to optimize dose distribution, then the quality of treatment plan can be improved, but the time and labor required for treatment planning increases
Solution Approach 1:
The system performs preliminary action by pre-calculating and storing optimal beam properties and dose distribution patterns in a database during the training phase. When a new treatment plan is needed, the system retrieves and applies pre-computed solutions rather than performing manual optimization, thus maintaining high dose distribution quality while dramatically reducing planning time.
Solution Approach 2:
The system creates copies of successful treatment plans and dose distribution patterns from the database that match the current patient's anatomical and clinical characteristics. These copied solutions serve as starting points or templates that can be quickly adapted, eliminating the need for time-consuming manual optimization while preserving the quality of professionally designed treatment approaches.
2Productivity
If inverse planning process is used to automate beam property adjustment, then treatment planning speed can be improved, but the ability to achieve optimal balance between target coverage and OAR sparing may be reduced
Solution Approach 1:
The system implements feedback by using the actual outcomes of treatment plans (dose distribution quality, OAR sparing effectiveness) to continuously improve the database of pre-computed solutions. The system learns from successful plans and refines its predictions, ensuring that automated planning not only maintains speed but also achieves optimal balance between target coverage and organ-at-risk sparing through iterative improvement.
Solution Approach 2:
The system performs self-service by autonomously selecting appropriate beam properties and dose distribution parameters from the database based on the patient's specific characteristics, without requiring manual intervention. The system automatically retrieves relevant pre-computed solutions, adapts them to the current case, and generates the treatment plan, thereby maintaining both speed and optimization quality through intelligent autonomous decision-making.
3Manufacturing precision
If achievable dose sparing goals are set for each OAR during inverse planning, then the overall plan quality can be improved, but the complexity of setting and balancing multiple goals increases
Solution Approach 1:
The system performs preliminary action by pre-determining and storing optimal dose sparing goals for various organ-at-risk structures in the database during the training phase. When creating a new treatment plan, the system automatically retrieves these pre-established goals rather than requiring clinicians to manually set and balance multiple competing objectives, thus maintaining high overall plan quality while eliminating the complexity of goal-setting.
Data Source
AI summary
Methods and systems for machine-learning-based prediction of a dose-volume histogram for radiotherapy treatment planning. The method includes receiving prescription information and a plan geometry. The plan geometry includes a planning target volume and an organ at risk. The method also includes extracting a plurality of input features using the plan geometry and a machine-learning model. The method further includes determining the dose-volume histogram by combining the plurality of input features using the machine-learning model.