Machine-Learned Fluence Maps for Radiation Therapy Planning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing radiation therapy planning methods are costly and time-consuming, particularly in inverse optimization processes, and struggle to translate two-dimensional image analysis to three-dimensional domains, complicating the delivery of radiation to tumors while minimizing exposure to healthy tissues.

Innovation Solution

A system utilizing machine learning models to generate fluence maps and leaf sequences directly, based on beam eye view projections and digitally reconstructed radiographs, reducing the need for iterative inverse planning by incorporating mechanical constraints of the linear accelerator.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If conventional inverse optimization techniques are used for treatment planning, then manufacturing precision (dose delivery accuracy) is improved, but productivity (planning speed) deteriorates

Engineering Contradiction:
Improvedose delivery accuracyVSAvoidplanning speed
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The patent replaces the conventional iterative inverse optimization computational system with a machine learning-based prediction system. The ML model is trained on historical treatment plans and directly predicts optimal fluence maps, eliminating the need for time-consuming iterative optimization calculations while maintaining dose delivery accuracy.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system performs preliminary action by pre-training the machine learning model on extensive historical treatment data before actual treatment planning. This pre-computed knowledge is then rapidly applied to new cases, avoiding the need for slow iterative optimization during the actual planning process.

Inventive Principle:
Principle #10Preliminary action

2Ease of operation

If knowledge-based planning techniques are used, then ease of operation is improved, but adaptability to three-dimensional domains deteriorates

Engineering Contradiction:
Improveplanning process simplicityVSAvoidthree-dimensional domain capability
Core Design Contradiction:
Ease of operationVSAdaptability or versatility

Solution Approach 1:

The patent transitions from two-dimensional image analysis to three-dimensional domain processing by using volumetric data from CT scans and generating three-dimensional fluence maps. The machine learning model processes 3D anatomical structures and delivers 3D dose distributions, maintaining simplicity while achieving full three-dimensional capability.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Manufacturing precision

If iterative inverse planning is performed, then manufacturing precision is improved, but loss of time increases

Engineering Contradiction:
Improvetreatment plan accuracyVSAvoidplanning time
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent substitutes the iterative inverse planning computational process with a direct machine learning prediction approach. The trained model instantly generates treatment plans with high accuracy, eliminating the repetitive iterative calculations that consume significant time while maintaining or improving plan quality.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP4613327A1Systems and methods for generating radiation therapy treatments plans
Publication Date: 2025.09.10 SIEMENS HEALTHINEERS INTERNATIONAL AG
  • EP4613327A1 patent drawingFigure 1
  • EP4613327A1 patent drawingFigure 2
  • EP4613327A1 patent drawingFigure 3A

AI summary

Provided herein are systems for planning radiotherapy treatments. Systems can include one or more processors to receive radiotherapy data associated with a set of treatments administered to a set of patients; generate a beam eye view (BEV) projection for each patient of the set of patients; and for each patient of the set of patients, provide treatment data associated with the patient and data associated with the BEV projections corresponding to the patient to a model to train the model to generate an output. The output can represent a fluence map. Systems and method for generating leaf sequences are also provided.