Radiotherapy Plan Optimization With Parallel Multi-Pass Dose Solving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional radiotherapy planning techniques require lengthy computation times and manual trial-and-error processes to optimize radiation dosage, especially when multiple anatomical regions are treated, necessitating two passes of optimization that are computationally intensive and resource-heavy.
Innovation Solution
Utilizing parallel processing hardware, specifically GPUs and ADMM techniques, to solve linear programming equations in multiple passes, optimizing radiation therapy plans by converting optimization problems into matrices and performing parallel executions to achieve faster and more accurate dose distributions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional sequential optimization methods are used for radiotherapy planning, then the optimization process can be performed with simple hardware, but the computation time becomes excessively long (from seconds to an hour per pass)
Solution Approach 1:
The optimization problem is segmented into multiple independent sub-problems that can be solved simultaneously on different processing units. Each pass optimizes a specific subset of dose points or weight combinations, allowing parallel execution that reduces total computation time from sequential to parallel processing
Solution Approach 2:
The patent introduces a temporal dimension to hardware resources by implementing multiple passes sequentially while using parallel processing within each pass. This allows the system to leverage parallel hardware capabilities without requiring permanent complex hardware configurations, achieving speedup through time-parallel execution
2Measurement precision
If multiple passes of optimization are performed to improve accuracy, then the dose distribution accuracy improves, but the total computation time and processing resources increase
Solution Approach 1:
The optimization process is divided into multiple passes, each handling a specific subset of dose points or weight combinations. By segmenting the workload, the system can perform multiple passes of optimization to improve accuracy while using parallel processing within each pass to reduce the time cost of each individual pass
Solution Approach 2:
The first pass performs preliminary optimization with a simplified geometric description of low-dose volumes, establishing initial dose points and weight combinations. This preliminary action provides a foundation for subsequent passes to refine the optimization, reducing the computational burden of later passes while maintaining improving accuracy
3Extent of automation
If manual trial-and-error is used to find optimal weight values, then the optimization process can be performed with minimal computational resources, but the process becomes manual and tedious
Solution Approach 1:
The system performs self-service optimization by automatically selecting optimal weight values and dose points without manual intervention. The optimization algorithm autonomously navigates the parameter space, evaluates different weight combinations, and identifies the optimal solution based on the treatment objectives, eliminating the manual trial-and-error process
Solution Approach 2:
The optimization process incorporates feedback mechanisms where the system evaluates the results of each optimization pass and uses this information to guide subsequent passes. The feedback from dose distribution calculations and convergence criteria allows the system to automatically adjust weight values and select optimal parameters without manual guidance
4Loss of time
If the first pass uses approximate geometric description of low-dose volumes, then the computation time is reduced, but the accuracy of low-dose volume control is insufficient
Solution Approach 1:
The first pass performs preliminary optimization using an approximate geometric description of low-dose volumes, quickly establishing initial dose points and weight combinations. This preliminary action with simplified geometry reduces computation time while still providing a foundation for the second pass to refine the low-dose volume control with more precise geometric descriptions
Solution Approach 2:
The optimization is segmented into two distinct passes: the first pass handles the computational efficiency with approximate geometry, while the second pass focuses on refining the low-dose volume control precision. This segmentation allows each pass to optimize for its specific goal, balancing time and precision requirements
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
Systems and methods are disclosed for optimization of radiotherapy treatments. Example operations for treatment planning include: obtaining first optimization problems for providing radiotherapy treatment to a human subject; performing dose optimization for delivery of the radiotherapy treatment to a treatment (target, low dose) area of the human subject, by performing at least a first pass and a second pass; and generating treatment plan data based on at least one of the multiple solutions provided by the second pass. In an example, the dose optimization includes: converting the first optimization problems into a first problem matrix; performing the first pass by solving the first optimization problems on parallel processing hardware; combining multiple solutions to the first optimization problems to produce second optimization problems; converting the second optimization problems into a second problem matrix; and performing the second pass by solving the second optimization problems on the parallel processing hardware.