MRI Contrast Assignment Using Neural Networks Across Protocol Variations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for assigning contrasts in magnetic resonance image datasets are prone to errors and manufacturer-specific, making them unreliable and inefficient, especially when protocol parameters change.
Innovation Solution
Utilizing artificial intelligence, specifically neural networks like ResNet, to automate the contrast assignment in magnetic resonance image datasets, enabling robust and manufacturer-independent classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual input or lookup tables are used for contrast assignment, then the process can be implemented with simple tools, but the reliability and accuracy of contrast assignment deteriorates due to errors and manufacturer-specific variations
Solution Approach 1:
The patent replaces manual or rule-based contrast assignment systems with an AI-based neural network system. The neural network automatically learns to identify and classify contrasts from MRI image data, eliminating the need for manual input or complex lookup tables while significantly improving reliability and accuracy in contrast assignment.
2Measurement precision
If lookup tables based on keywords or metainformation are used, then the system can operate automatically, but measurement precision deteriorates due to textual data differences between software versions and manufacturers
Solution Approach 1:
The patent changes the approach from using textual parameters (keywords, metainformation) to using neural network models that process image data directly. This allows the system to adapt to different software versions and manufacturers by learning patterns from actual image data rather than relying on potentially inconsistent textual descriptions.
Solution Approach 2:
The neural network system performs self-learning and self-adjustment to identify contrasts automatically without requiring manual intervention or predefined lookup tables. The model continuously adapts to different data formats and manufacturers by training on diverse MRI datasets, making it inherently versatile across different software versions and hardware systems.
3Adaptability or versatility
If protocol parameters or sequence parameters are changed, then the imaging flexibility and versatility improve, but the reliability of contrast assignment deteriorates when text in protocol names is not adapted
Solution Approach 1:
The patent replaces text-based protocol name matching with a neural network system that directly analyzes image characteristics. This substitution allows the system to maintain high reliability even when protocol parameters change, as the neural network learns to identify contrasts based on actual image data patterns rather than relying on potentially outdated or mismatched textual protocol names.
Data Source
AI summary
In a computer-implemented method for assigning at least one contrast to at least one magnetic resonance image dataset of a recording region of a patient, input data comprising magnetic resonance data of the magnetic resonance image dataset is received, at least one trained assigning function is applied to the input data, and the output data is provided. The output data includes a contrast assigning information item. The contrast exists due to the use of a contrast-specific magnetic resonance sequence during the recording.