Radiotherapy Treatment Planning Using Machine Learning Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current radiotherapy treatment planning is time-consuming, resource-intensive, and subjective, often relying on dose-volume histograms that lack spatial information and require intensive calculations, leading to suboptimal treatment plans.
Innovation Solution
A method and system using machine learning techniques to generate treatment plans by determining a training model from feature and target vectors, allowing for the creation of a therapy model that optimizes radiation distribution based on clinical and dosimetric objectives, ensuring maximum dose to the tumor while minimizing dose to surrounding healthy tissue.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If automatic treatment planning using reference plans and dose-volume histograms is employed, then treatment planning time is reduced, but the quality and accuracy of treatment plans deteriorates due to loss of spatial information
Solution Approach 1:
The patent creates a digital twin (virtual copy) of the patient's anatomy with preserved spatial relationships. Instead of using compressed dose-volume histograms that lose spatial information, the system copies the full 3D anatomical structure and dose distribution data from reference patients, maintaining all spatial details while enabling automated planning.
Solution Approach 2:
The patent transitions from 1D dose-volume histogram representation to 3D spatial representation by creating a virtual patient model that preserves full spatial dimensions. This dimensional upgrade allows the system to maintain spatial information while still enabling automated processing through the digital twin framework.
2Manufacturing precision
If voxel-based optimization (dose painting) is used to specify different criteria for every voxel, then treatment plan accuracy is improved, but computational complexity and resource requirements increase dramatically
Solution Approach 1:
The virtual patient model automatically generates optimized dose distributions by leveraging the embedded spatial relationships and anatomical structures. Instead of requiring complex iterative optimization algorithms to process millions of voxels, the system uses the pre-processed 3D model to directly determine optimal dose patterns, making the system self-sufficient and computationally efficient.
Solution Approach 2:
The patent performs preliminary processing by creating the 3D virtual patient model with all spatial relationships established beforehand. This pre-computation of anatomical structures and their spatial relationships eliminates the need for complex real-time optimization calculations during treatment planning, significantly reducing computational complexity while maintaining high accuracy.
3Extent of automation
If dose-volume histograms are used to match reference plans, then treatment planning becomes more automated, but the optimization process becomes more difficult and requires intensive iterative calculations
Solution Approach 1:
Instead of using compressed dose-volume histograms that require intensive iterative optimization, the patent copies the complete 3D dose distribution data from reference patients into the virtual patient model. This copying approach maintains all spatial information and eliminates the need for complex iterative DVH matching calculations.
Solution Approach 2:
The patent inverts the traditional approach by not trying to match DVH curves through iterative optimization, but instead directly copying and adapting 3D dose distributions from reference cases. This inversion transforms a complex optimization problem into a simpler data transfer and adaptation process.
Data Source
AI summary
The present disclosure relates to systems, methods, and computer-readable storage devices for radiotherapy treatment planning. For example, a method may generate a treatment plan for a patient. The method may receive training data reflecting radiotherapy treatment data. The training data may include a feature vector and a target vector. The method may further determine a training model based on the feature vector and the target vector. The method may further receive testing data associated with the patient. The testing data may include a descriptive feature vector. The method may further determine a therapy model based on the descriptive feature vector and the training model. The therapy model may be used to generate the treatment plan.