Machine-Learned Collimator Settings for Interventional X-Ray
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current X-ray imaging systems require cumbersome and time-consuming manual adjustments of the collimator, which are often not optimally set and need re-adjustment during interventions, leading to increased radiation exposure and procedure duration.
Innovation Solution
A system utilizing a machine learning model to estimate collimator settings based on user input and acquired images, allowing for quick and adaptive collimator adjustments through a collimator setting estimator that computes a complete setting from partial user input.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If manual collimator adjustment is performed, then collimator settings can be changed, but the process becomes time-consuming and cumbersome
Solution Approach 1:
The system uses machine learning models to automatically compute optimal collimator settings based on input images and user preferences, eliminating the need for manual adjustment of multiple collimator components. The algorithm self-determines the appropriate field of view and collimator positions without requiring user intervention in the adjustment process.
Solution Approach 2:
The system pre-computes collimator settings by analyzing input images and determining optimal field of view parameters before the actual imaging procedure begins. This preliminary determination of collimator positions and angles eliminates the need for time-consuming adjustments during the procedure.
2Object-affected harmful factors
If tight collimation is applied to reduce radiation dose, then patient and user exposure is reduced, but the collimator settings become less adaptable when switching imaging geometries
Solution Approach 1:
The system dynamically adapts collimator settings based on the current imaging geometry and region of interest. When the imaging geometry changes or the user switches to a different ROI, the machine learning algorithm automatically recomputes optimal collimator parameters, maintaining tight collimation for radiation reduction while ensuring adaptability to new imaging configurations.
Solution Approach 2:
The system incorporates user feedback and imaging geometry information to continuously optimize collimator settings. The algorithm monitors the current imaging configuration and adjusts collimator parameters in real-time to maintain optimal radiation reduction while preserving the ability to adapt to changing imaging requirements.
3Manufacturing precision
If multiple collimator components are adjusted manually, then precise collimation can be achieved, but the complexity of operation increases
Solution Approach 1:
The machine learning system serves as a universal controller that manages all collimator components (shutters and wedges) through a single integrated algorithm. Instead of requiring separate manual adjustment of multiple components, the system computes and controls all collimator settings through one automated process, maintaining precision while reducing operational complexity.
Solution Approach 2:
The machine learning algorithm acts as an intermediary between the user's imaging requirements and the physical collimator components. The algorithm translates high-level imaging goals into specific collimator settings for multiple components, eliminating the need for users to directly manage the complexity of individual shutter and wedge adjustments while maintaining precise collimation.
Data Source
Figure 1
Figure 2A~2C
Figure 3
AI summary
A system (SYS) and related method for facilitating collimator adjustment in X-ray imaging or a radiation therapy delivery. The system comprises an input interface (IN) for receiving input data including i) an input image and/or ii) user input data including a partial collimator setting for a collimator (COL) of an X-ray imaging apparatus (IA). A collimator setting estimator (CSE) of the system computes a complemented collimator setting for the collimator based the input data. Preferably, the system uses machine learning.