Machine Learning Parameter Estimation for MRI Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional MR data processing techniques face challenges in efficiently estimating starting parameters for non-linear regression procedures, leading to increased processing time and the risk of imaging artefacts, especially in biological tissues where different tissue types and field conditions complicate initial parameter estimation.
Innovation Solution
A method and apparatus that utilize a machine learning-based estimation procedure to provide input parameters for a multi-parameter non-linear regression analysis of MR data, incorporating models derived from Bloch equations, allowing for fast and reliable parameter mapping with inherent verification of input parameters through goodness-of-fit parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional non-linear regression procedures are used for MR data processing, then parameter estimation can be performed with model-based verification, but processing time increases and the risk of incorrect parameter estimation arises due to difficulty in finding appropriate starting parameters
Solution Approach 1:
The patent applies machine learning models to pre-estimate starting parameters before the non-linear regression procedure begins. This preliminary action provides the regression algorithm with optimized initial values, eliminating the time-consuming trial-and-error process of finding appropriate starting parameters while ensuring accurate convergence to the correct solution.
Solution Approach 2:
The patent introduces machine learning models as an intermediary between the raw MR data and the non-linear regression procedure. This intermediary layer processes the data to extract meaningful starting parameters, bridging the gap between raw measurements and the complex regression analysis, thereby reducing both time and computational complexity.
2Productivity
If machine learning based estimation procedure is used to provide input parameters, then processing time is reduced, but model-based verification capability may be compromised
Solution Approach 1:
The patent merges machine learning-based parameter estimation with model-based non-linear regression in a hybrid approach. The machine learning component provides fast initial parameter estimates, while the subsequent regression procedure with model-based verification refines these estimates and provides reliability assessment through goodness-of-fit parameters, combining the speed of ML with the reliability of physical models.
Solution Approach 2:
The patent implements a feedback mechanism where the non-linear regression procedure evaluates the machine learning estimates using model-based goodness-of-fit parameters. This feedback loop verifies the reliability of ML-generated parameters and can adjust or re-estimate parameters that do not meet the model-based criteria, ensuring both speed and reliability.
3Measurement precision
If manual estimation of starting parameters is performed for each subject, then parameter accuracy can be optimized, but time consumption increases significantly
Solution Approach 1:
The patent enables the system to automatically estimate its own starting parameters using machine learning models trained on MR data characteristics. This self-service capability eliminates the need for manual parameter estimation by operators, providing accurate starting values automatically for each subject while maintaining high processing throughput through automated ML inference.
Solution Approach 2:
The patent transforms the parameter estimation task from a manual, subject-specific optimization process into an automated parameter change detection problem. The machine learning model learns to identify characteristic parameter patterns across different subjects and tissue types, automatically adapting parameters based on data characteristics rather than manual adjustment.
Data Source
AI summary
A method of processing magnetic resonance (MR) data of a sample under investigation, includes the steps of providing the MR data being collected with an MRI scanner apparatus, and subjecting the MR data to a multi-parameter nonlinear regression procedure being based on a non-linear MR model and employing a set of input parameters, wherein the regression procedure results in creating a parameter map of model parameters of the sample, wherein the input parameters (initial values and possibly boundaries) of the regression procedure are estimated by a machine learning based estimation procedure applied to the MR data. The machine learning based estimation procedure preferably includes at least one of at least one neural network and a support vector machine. Furthermore, an MRI scanner apparatus is described.