AI Temporal Prediction for Radiotherapy Anatomic Motion Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current imaging techniques for radiotherapy are unable to accurately track patient motion in real-time due to limitations in temporal resolution, leading to incomplete or incorrect anatomic position monitoring, which affects the precision of radiation therapy delivery.
Innovation Solution
Utilizing a specially trained Transformer AI model to estimate respiratory motion from prior breathing cycles, enabling continuous tracking of 3D positions of target regions during radiotherapy, allowing for real-time adjustments to the radiation beam.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional imaging techniques (3D CBCT or 3D MRI) are used to monitor patient motion, then volumetric 3D images can be acquired, but the scan time is too long to capture respiratory motion with sufficient temporal resolution
Solution Approach 1:
The patent uses 2D images as copies or projections of the 3D anatomy to estimate motion. Instead of acquiring full 3D volumetric images continuously, the system uses multiple 2D projections (which are faster to acquire) and reconstructs the 3D motion information through image registration and transformation algorithms. This copying approach maintains measurement precision while dramatically reducing scan time.
Solution Approach 2:
The patent replaces the mechanical imaging acquisition system with an computational estimation system. Rather than physically scanning the entire volume repeatedly (mechanical approach), the system uses 2D image registration, feature tracking, and transformation algorithms (computational approach) to estimate 3D position changes, thereby reducing acquisition time while maintaining monitoring accuracy.
2Productivity
If surface information (sensors on patient or markers on vest) is used to estimate patient motion, then real-time motion data can be obtained, but the assumption that surface information correlates to internal patient state is often inaccurate
Solution Approach 1:
The patent introduces 2D imaging data and image registration algorithms as intermediaries between surface markers and internal anatomy. Instead of directly assuming correlation between surface motion and internal motion, the system uses 2D images captured during treatment to establish transformation relationships that more accurately reflect the actual 3D anatomical position changes, thereby improving measurement precision while maintaining real-time capability.
Solution Approach 2:
The patent implements a feedback loop where 2D images are continuously acquired during treatment, registered to reference images, and used to update motion estimates in real-time. This feedback mechanism allows the system to correct for discrepancies between surface marker motion and internal anatomy motion, improving the accuracy of anatomic position monitoring while maintaining real-time productivity.
3Object-affected harmful factors
If treatment margins are reduced to improve radiation therapy precision, then exposure to unintended radiation decreases, but the ability to accurately track moving targets becomes more challenging
Solution Approach 1:
The patent implements continuous 2D image acquisition and real-time motion estimation throughout the radiation treatment process. This continuous monitoring allows the system to track target position changes dynamically, enabling reduced treatment margins while maintaining reliable target tracking. The system continuously updates motion estimates and can trigger beam gating or adjustment based on real-time position feedback.
Solution Approach 2:
The patent enables the treatment system to self-adjust based on real-time motion estimation. The system automatically registers images, calculates transformation parameters, and triggers beam gating or position adjustment without requiring manual intervention. This self-service capability ensures reliable target tracking even with reduced margins, as the system autonomously compensates for motion in real-time.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
Systems and methods are disclosed for monitoring and estimating an anatomic position of a human subject for a radiotherapy treatment session, based on use of an artificial intelligence (AI) model (e.g., a generative AI model comprising a Transformer deep learning neural network), are described. An example method of monitoring anatomic position with a trained AI model includes: receiving position information corresponding to observed positions of a tracked anatomical area of a patient, observed during the radiotherapy treatment session; providing the position information as an input to a trained model trained with temporal sequences of observed anatomical positions from training data; determining an estimated position of the tracked anatomical area of the patient at a future time, based on output of the trained model; and controlling the radiotherapy treatment session based on the estimated position of the tracked anatomical area of the patient.