Radiation Treatment Planning With ML Isocenter Placement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for treatment planning in radiation therapy, particularly for systems like the Leksell Gamma Knife, are inefficient and time-consuming, often leading to sub-optimal distribution of shot sizes and locations due to poor dose distribution descriptions, requiring excessive computational resources and manual intervention.
Innovation Solution
A machine learning-based approach for treatment planning that utilizes supervised or unsupervised models, such as neural networks, to estimate isocenter locations and beam-on times, optimizing a treatment quality criterion in two phases to generate efficient geometric configurations for radiation therapy plans.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If manual treatment planning is performed by clinical experts, then treatment plan quality can be optimized, but the process becomes extremely time-consuming and challenging due to many interdependent decisions
Solution Approach 1:
The patent replaces the manual mechanical process of treatment planning with an automated computational system using machine learning models. The system automatically determines isocenter locations and shot parameters by processing target volume geometry and applying trained algorithms, eliminating the need for manual expert intervention while maintaining or improving plan quality.
Solution Approach 2:
The patent introduces machine learning models as intermediaries between the target volume geometry and the treatment plan parameters. These models serve as a bridge that automatically translates geometric information into optimized isocenter locations and shot configurations, resolving the complexity of interdependent decisions without manual intervention.
2Productivity
If automated isocenter selection methods are used, then planning efficiency is improved, but the dose distribution description becomes poor leading to sub-optimal shot size distribution
Solution Approach 1:
The patent applies preliminary action by training machine learning models on extensive datasets of treatment plans and dose distributions before actual planning. This pre-training allows the model to learn optimal relationships between target geometry and dose distribution characteristics, enabling it to predict isocenter locations that inherently produce superior shot size distributions without requiring complex real-time optimization.
Solution Approach 2:
The patent changes the approach from geometric parameter optimization to data-driven parameter prediction. Instead of manually adjusting geometric parameters to achieve optimal dose distribution, the system uses machine learning models that have learned optimal parameter configurations from training data, automatically selecting isocenter locations that produce superior dose distributions.
3Loss of energy
If conventional optimization methods are used, then computational resources can be managed, but the treatment planning process remains inefficient and requires excessive manual intervention
Solution Approach 1:
The patent replaces conventional iterative optimization algorithms with machine learning-based prediction models. This substitution eliminates the need for computationally intensive iterative optimization processes while dramatically improving planning efficiency. The trained models directly predict optimal isocenter locations without requiring repeated computational iterations.
Solution Approach 2:
The patent uses copying by training the machine learning model on extensive datasets of existing treatment plans and their corresponding optimal solutions. The model learns from these copied examples and generalizes the knowledge to new cases, achieving high efficiency without requiring computationally expensive real-time optimization for each new patient.
Data Source
AI summary
The present disclosure relates to the field of radiation therapy and methods, software and systems for treatment planning.