Transfer Learning for Faster, Consistent Radiotherapy Planning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional radiotherapy treatment planning is time and labor intensive, requiring skilled professionals and is prone to variations in structure delineation, leading to uncertainties in target volume and OAR radiation doses.

Innovation Solution

Deep transfer learning is applied to generate a target deep learning engine by configuring variable or invariable base layers of a pre-trained base deep learning engine, utilizing base training data to adapt the engine for specific radiotherapy tasks, reducing the need for extensive user expertise and computational resources.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional radiotherapy treatment planning is performed manually by skilled professionals, then treatment planning accuracy is maintained, but time consumption and labor intensity increase significantly

Engineering Contradiction:

Improvestructure delineation accuracyVSAvoidtreatment planning time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces manual mechanical delineation work with an automated deep learning system. The neural network model automatically segments anatomical structures and generates treatment plans, substituting the manual mechanical process of skilled professionals with an automated computational system that maintains accuracy while dramatically reducing time consumption.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system enables self-service treatment planning through the automated deep learning engine. The model processes patient imaging data and generates treatment plans autonomously without requiring continuous human intervention, allowing the system to serve itself in generating initial treatment plans that can be reviewed and approved by clinicians.

Inventive Principle:

Principle #25Self-service

2Manufacturing precision

If manual structure delineation is performed to ensure accurate target volume and OAR definition, then treatment planning precision is improved, but variability and uncertainty increase due to human factors

Engineering Contradiction:

Improvetarget volume definition precisionVSAvoidconsistency of structure delineation

Core Design Contradiction:

Manufacturing precisionVSReliability

Solution Approach 1:

The patent changes the fundamental parameter of who performs the delineation from human operators to an automated neural network. This parameter change eliminates human variability and consistency issues while maintaining or improving precision. The deep learning model applies the same segmentation criteria uniformly across all cases, removing the inter-observer and intra-observer variability inherent in manual delineation.

Inventive Principle:

Principle #35Parameter changes

3Productivity

If deep transfer learning is applied to generate target deep learning engines, then productivity and efficiency are improved, but system complexity increases

Engineering Contradiction:

Improvetreatment planning throughputVSAvoiddeep learning engine architecture

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by pre-training deep learning engines on large datasets of base radiotherapy treatment planning cases before deploying them for target tasks. This pre-training establishes a foundation that can be adapted to specific clinical scenarios, improving productivity while managing complexity through transfer learning rather than training from scratch for each application.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The deep learning engine architecture is designed with universality in mind, where a single base engine can be adapted through transfer learning to perform multiple different radiotherapy treatment planning tasks. This multi-functionality approach reduces overall system complexity by using a common foundation model rather than requiring separate specialized models for each task.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS12580063B2Methods and systems for radiotherapy treatment planning based on deep transfer learning

Publication Date: 2026.03.17 SIEMENS HEALTHINEERS INTERNATIONAL AG

AI summary

Example methods and systems for deep transfer learning for radiotherapy treatment planning are provided. One example method may comprise: obtaining (310) a base deep learning engine that is pre-trained to perform a base radiotherapy treatment planning task; and based on the base deep learning engine, generating a target deep learning engine to perform a target radiotherapy treatment planning task. The target deep learning engine may be generated by configuring (330) a variable base layer among multiple base layers of the base deep learning engine, and generating (340) one of multiple target layers of the target deep learning engine by modifying the variable base layer. Alternatively or additionally, the target deep learning engine may be generated by configuring (350) an invariable base layer among the multiple base layers, and generating (360) one of multiple target layers of the target deep learning engine based on feature data generated using the invariable base layer.