Methods and systems for radiotherapy treatment planning using deep learning engines
Inventors
Laaksonen, Hannu • CORDERO MARCOS, María • CZEIZLER, Elena • Nord, Janne • PERTTU, Sami Petri
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Assignees
Siemens Healthineers International AG
Member
Siemens HealthineersSiemens Healthineers is dedicated to pioneering breakthroughs in healthcare, providing innovative technologies and services in diagnostic and therapeutic imaging, laboratory diagnostics, and digital health solutions. Our mission is to enhance healthcare providers' value and improve patient outcomes globally. We are also committed to sustainability, aiming to minimize environmental impact while advancing healthcare.
Siemens Healthineers is dedicated to pioneering breakthroughs in healthcare, providing innovative technologies and services in diagnostic and therapeutic imaging, laboratory diagnostics, and digital health solutions. Our mission is to enhance healthcare providers' value and improve patient outcomes globally. We are also committed to sustainability, aiming to minimize environmental impact while advancing healthcare.
Publication Number
US-12353989-B2
Publication Date
2025-07-08
Expiration Date
2038-09-28
Abstract
Example methods for radiotherapy treatment planning using deep learning engines are provided. One example method may comprise obtaining first image data associated with a patient; generating first feature data by processing the first image data associated with a first resolution level using a first processing pathway; generating second feature data by processing second image data associated with a second resolution level using a second processing pathway; and generating third feature data by processing third image data associated with a third resolution level using a third processing pathway. The example method may also comprise generating a first combined set of feature data associated with the second resolution level, and a second combined set of feature data associated with the first resolution level based on the first feature data and the first combined set. Further, the example method may comprise generating output data associated with radiotherapy treatment of the patient.
Core Innovation
The invention provides methods and systems for radiotherapy treatment planning using a deep learning engine that includes multiple processing pathways to process medical image data at different resolution levels. The deep learning engine processes first image data associated with a first resolution level using a first processing pathway, second image data at a second resolution level using a second processing pathway, and third image data at a third resolution level using a third processing pathway. By combining feature data generated from these pathways at different resolution levels, the system generates output data associated with radiotherapy treatment planning, which may include structure data, dose data, or treatment delivery data.
The problem addressed is that conventional radiotherapy treatment planning and adaptive radiotherapy treatment planning are time and labor intensive, requiring highly skilled experts to manually delineate structures on image data, perform segmentation, dose prediction, and generate treatment plans. There is significant variability and uncertainty in manual segmentation, which can impact treatment accuracy. The invention aims to automate these steps using deep learning to improve efficiency, accuracy, and adaptiveness in radiotherapy treatment planning.
Additionally, the invention addresses adaptive radiotherapy treatment planning challenges where the patient's anatomical conditions may differ from the planning images during treatment. It provides methods to update treatment plans based on treatment image data acquired during the treatment phase, either by transforming treatment images to match planning image modalities or by jointly processing multiple image modalities using deep learning. This supports timely and accurate updates of treatment plans to improve treatment delivery quality.
Claims Coverage
The patent claims cover three independent claims directed to a method, a non-transitory computer-readable medium, and a computer system, each for performing radiotherapy treatment planning using a deep learning engine with three processing pathways.
Multi-resolution processing by multiple pathways
Generating first, second, and third feature data by processing respective image data at first, second, and third resolution levels using first, second, and third processing pathways of a deep learning engine.
Combined feature data generation across resolution levels
Generating a first combined set of feature data at the second resolution level based on second and third feature data, and generating a second combined set of feature data at the first resolution level based on first feature data and the first combined set.
Use of convolution layers for feature extraction and combination
Processing image data using sets of convolution layers within each processing pathway and convolution layers for combining upsampled or downsampled feature data.
Resolution adjustments between pathways
Downsampling image data to generate lower resolution inputs for subsequent pathways and upsampling feature data to combine feature data across different resolution levels.
Output generation for radiotherapy treatment planning
Generating output data associated with radiotherapy treatment of a patient based on the combined feature sets, via one or more mixing layers of the deep learning engine.
Training for multiple radiotherapy planning tasks
Training the deep learning engine using past patient data to perform one or more of automatic segmentation (structure data output), dose prediction (dose data output), or treatment delivery data estimation (treatment delivery data output).
Together, the claims cover a multi-pathway deep learning-based radiotherapy planning system that separately processes image data at different resolutions, combines feature data in stages, and generates output data useful for treatment planning tasks such as segmentation, dose prediction, and treatment delivery estimation.
Stated Advantages
Improves efficiency of radiotherapy treatment planning by automating segmentation, dose prediction, and treatment delivery data estimation using deep learning.
Enhances accuracy and reliability in planning by leveraging multi-resolution image processing pathways to capture both local and global anatomical features.
Reduces manual labor and inter-physician variability associated with manual delineation and planning.
Enables adaptive radiotherapy planning by quickly updating treatment plans based on treatment-phase imaging data, thus potentially improving treatment outcomes and patient satisfaction.
Allows use of multiple imaging modalities and image registrations to improve model inputs and treatment plan adaptation quality.
Documented Applications
Automatic segmentation of medical images to generate structure data (contours and labels of target and organs at risk).
Dose prediction for targets and organs at risk in radiotherapy treatment planning.
Treatment delivery data estimation including beam orientations, dose fluence maps, machine control point data for radiotherapy systems.
Adaptive radiotherapy treatment planning (ART) to update treatment plans based on intra-treatment imaging data such as CBCT and CT images.
Transformation of treatment image data to planning image modality (e.g., CBCT to synthetic CT) for ART using deep learning or registration approaches.
Joint processing of multiple image modalities (e.g., CBCT and CT) to generate output data for ART planning updates.
Interested in licensing this patent?