ML Fluence Map Prediction for Automated Radiation Treatment Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional radiation treatment planning in IMRT requires significant human intervention and time due to the need for manual adjustment of fluence maps through inverse optimization, which is inefficient and prone to inconsistencies.
Innovation Solution
A machine learning system that predicts fluence maps directly from patient data, including CT scans and physician prescriptions, using convolutional neural networks and generative adversarial networks, eliminating the need for inverse optimization and enabling automated generation of radiation treatment plans.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional inverse optimization is used for IMRT treatment planning, then dosimetric quality can be achieved, but planning time is significantly increased and consistency is reduced
Solution Approach 1:
The patent replaces the conventional inverse optimization mechanical system with a machine learning-based fluence map prediction system. The ML model directly predicts optimal fluence maps from patient anatomy and treatment parameters, eliminating the iterative trial-and-error optimization process while maintaining dosimetric quality and significantly reducing planning time.
Solution Approach 2:
The patent performs preliminary action by pre-training the machine learning model on extensive datasets of treatment plans and their corresponding fluence maps. This pre-learning phase enables the system to rapidly generate high-quality fluence maps during actual treatment planning without requiring time-consuming iterative optimization for each new patient case.
2Manufacturing precision
If manual adjustment of fluence maps is performed, then treatment plan quality can be improved, but the process becomes highly time-consuming and prone to inconsistencies
Solution Approach 1:
The patent implements self-service by enabling the machine learning system to automatically generate and optimize fluence maps without requiring manual adjustment by radiation therapists. The system learns from historical data and autonomously produces high-quality treatment plans, eliminating human variability and significantly improving planning efficiency while maintaining consistent quality.
3Measurement precision
If iterative communication between planners and radiation oncology team members is conducted, then treatment plan accuracy can be improved, but the process becomes highly time-consuming
Solution Approach 1:
The patent incorporates feedback mechanisms where the machine learning system continuously learns from treatment plan outcomes and dosimetric results. The system uses feedback from dosimetry calculations and treatment plan validation to refine fluence map predictions, achieving high accuracy without requiring iterative communication cycles between team members.
Data Source
AI summary
A radiation treatment planning system can include a machine learning system that receives patient data, including an image scan (e.g., CT scan) and contour(s), a physician prescription, including planning target and dose, and device (radiation beam) data and outputs predicted fluence maps. The machine learning system includes at least two stages, where a stage of the at least two stages includes converting image scans from the patient data to projection images. A treatment planning system can receive the predicted fluence maps and generates treatment plans without performing inverse optimization.