ML-Based Acquisition Parameter Selection for Medical Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The selection of appropriate acquisition parameters for medical imaging systems, such as X-ray, Magnetic Resonance, and Ultrasound, is often challenging due to the need for patient-specific configurations, leading to potential increased radiation exposure, longer examination times, and insufficient image quality, as current methods rely heavily on landmark detection and operator expertise, which may not account for all relevant patient information.
Innovation Solution
A system and method that utilize a machine learning algorithm to identify acquisition parameters by analyzing depth-related maps generated from camera sensor data, incorporating both depth and non-image patient data, such as weight, age, and disease diagnoses, to model the relationship between patient attributes and imaging parameters, thereby providing a more holistic approach that does not solely rely on landmark detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If landmark detection is used to determine imaging configuration, then the imaging system can be configured based on detected landmarks, but not all relevant landmarks may be reliably detected and other potentially useful information in the depth image is not used
Solution Approach 1:
The depth image is used for multiple purposes: both for landmark detection and for feeding into the machine learning algorithm for acquisition parameter selection. This multi-functional use ensures that all available depth information is utilized effectively, rather than being limited to landmark detection only.
Solution Approach 2:
The system transitions from relying solely on landmark detection parameters to using a machine learning algorithm that processes the entire depth image. This parameter change enables the system to consider all relevant information in the depth image, not just detected landmarks, for determining optimal imaging configuration.
2Ease of operation
If a technician manually configures acquisition parameters, then expertise and experience can be applied, but trial and error may result in increased radiation exposure and longer examination time
Solution Approach 1:
The system enables self-service by automatically selecting acquisition parameters using a machine learning algorithm that processes depth image information. This eliminates the need for technician trial and error, reducing examination time while maintaining or improving parameter selection quality.
Solution Approach 2:
The machine learning algorithm is trained on previously annotated depth images and corresponding optimal acquisition parameters. This feedback mechanism allows the system to learn from past cases and automatically apply optimal parameters to current patients, eliminating manual trial and error processes.
3Reliability
If collimation is increased to ensure sufficient coverage, then diagnostic value is improved, but over-collimation increases radiation exposure to the patient
Solution Approach 1:
The system dynamically adjusts collimation parameters based on depth image analysis and machine learning predictions. By considering the entire depth image rather than just landmarks, the system can precisely determine the required coverage area, optimizing collimation to provide sufficient diagnostic coverage while minimizing radiation exposure to areas outside the region of interest.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
A system and method are provided for selecting an acquisition parameter for an imaging system. The acquisition parameter at least in part defines an imaging configuration of the imaging system during an imaging procedure with a patient. A depth- related map is accessed which is generated on the basis of sensor data from a camera system, wherein the camera system has a field of view which includes at least part of a field of view of the imaging system, wherein the sensor data is obtained before the imaging procedure with the patient and indicative of a distance that parts of the patient's exterior have towards the camera system. A machine learning algorithm is applied to the depth-related map to identify the acquisition parameter, which may be provided to the imaging system.