AI Dose Deposition Correction for Fast Radiation Planning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing radiation therapy dose calculation methods, particularly for small tumors, suffer from inaccuracies due to suboptimal modeling of radiation behavior, leading to difficulties in creating precise treatment plans that minimize damage to healthy tissue while effectively targeting cancerous tumors.

Innovation Solution

An artificial intelligence model, such as a neural network, is trained to enhance the accuracy of dose deposition calculations by learning the relationships between different dosing algorithms, allowing it to mimic the results of more computationally intensive methods like Monte Carlo simulations without the need for their full execution.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If Monte Carlo simulations are used to calculate dose deposition, then measurement precision is improved, but productivity deteriorates due to slow computation speed

Engineering Contradiction:
Improvedose deposition accuracyVSAvoidcomputation speed
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The system performs preliminary actions by pre-calculating and storing correction factors through Monte Carlo simulations during an offline training phase. These pre-computed correction factors are then applied during clinical dose calculations, eliminating the need to run full Monte Carlo simulations for each patient while maintaining high accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system creates a simplified copy or approximation of the complex Monte Carlo simulation by training a neural network model to replicate its results. The neural network learns the relationship between input parameters and Monte Carlo output, enabling fast dose calculations that mimic Monte Carlo accuracy without the computational burden.

Inventive Principle:
Principle #26Copying

2Productivity

If deterministic methods are used to calculate dose deposition, then productivity is improved due to faster computation, but measurement precision deteriorates

Engineering Contradiction:
Improvecomputation speedVSAvoiddose deposition accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The system introduces an intermediary neural network model that bridges deterministic methods and Monte Carlo simulations. The neural network takes deterministic calculation results as input and applies learned corrections to achieve Monte Carlo-level accuracy, acting as a mediator between the two approaches.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system changes the parameters of deterministic calculations by applying correction factors that adjust the results toward Monte Carlo accuracy. These correction factors are derived from Monte Carlo data and modify key parameters such as dose values and distribution patterns to improve precision.

Inventive Principle:
Principle #35Parameter changes

3Ease of operation

If empirical models are used for small tumors, then ease of operation is improved, but measurement precision deteriorates due to model generality

Engineering Contradiction:
Improvetreatment planning simplicityVSAvoiddose deposition accuracy
Core Design Contradiction:
Ease of operationVSMeasurement precision

Solution Approach 1:

The system applies local quality by training separate neural network models for different tumor sizes, anatomical locations, and treatment types. Each model is specialized for specific conditions, providing locally optimized accuracy rather than relying on a single general-purpose empirical model.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4489854B1Artificial intelligence boosting dose calculations
Publication Date: 2026.02.18 SIEMENS HEALTHINEERS INTERNATIONAL AG
  • EP4489854B1 patent drawingFigure 1
  • EP4489854B1 patent drawingFigure 2
  • EP4489854B1 patent drawingFigure 3

AI summary

Embodiments described herein provide for training an artificial intelligence model (200, 300) to boost dose depositions. The artificial intelligence model (200, 300) receives medical images and a dose deposition determined according to a first dose deposition model. The artificial intelligence model (200, 300) modifies the received dose deposition determined according to the first dose deposition model such that the dose deposition simulates a dose deposition determined by a second dose deposition model.