Deep Learning Eddy Current Correction for MRI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for correcting eddy currents in MRI images require tedious manual processing by radiologists, leading to potential artifacts and misrepresentations in flow velocity measurements, especially in cardiac scans, necessitating an automated solution.
Innovation Solution
A deep learning-based convolutional neural network (CNN) model is trained to automatically generate eddy current correction masks for 2D and 4D MR images, using annotated data and post-processing techniques to improve segmentation accuracy and reduce radiologist intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual eddy current correction methods are used with interpolation functions and ECC masks, then eddy current effects can be corrected, but the process requires tedious manual processing by radiologists and extensive pre-processing
Solution Approach 1:
The system performs self-service by automatically generating ECC masks through deep learning without requiring radiologist intervention. The CNN model processes the MRI data independently, identifying static and non-static tissue regions and generating correction masks autonomously, thereby eliminating the tedious manual processing step while maintaining correction accuracy.
Solution Approach 2:
The manual mechanical process of radiologist-based mask generation and adjustment is replaced with an automated computational system. The deep learning model substitutes the human-in-the-loop workflow with algorithm-based automatic segmentation and mask generation, dramatically reducing processing time while preserving the essential correction function.
2Reliability
If intensity binning and threshold adjustment methods are used to generate ECC masks, then eddy current correction can be applied, but the process requires arbitrary manual work outside the scope of diagnosis
Solution Approach 1:
The system eliminates the need for radiologist intervention in threshold adjustment and mask refinement. The CNN model automatically determines appropriate thresholds and generates accurate ECC masks through learned patterns from training data, making the system self-sufficient and removing arbitrary manual work from the diagnostic workflow.
Solution Approach 2:
The model performs preliminary action by pre-learning optimal thresholding and segmentation strategies during the training phase. This preliminary training on annotated data enables the model to automatically perform what would otherwise require manual threshold adjustment, streamlining the operational workflow while maintaining mask accuracy.
3Extent of automation
If automated deep learning-based CNN models are used to generate ECC masks, then radiologist intervention is reduced and processing is automated, but the system requires training data and model development
Solution Approach 1:
The complexity of model development and training is performed as a preliminary action before deployment. The CNN model is trained offline on annotated MRI data to learn optimal features and parameters. Once trained, the model can be deployed for automated ECC mask generation without requiring further manual intervention or complex runtime adjustments, separating the complexity of development from the simplicity of operation.
Data Source
AI summary
Systems and methods for providing improved eddy current correction (ECC) in medical imaging environments. One or more of the embodiments disclosed herein provide a deep learning-based convolutional neural network (CNN) model trained to automatically generate an ECC mask which may be composited with two-dimensional (2D) scan slices or four-dimensional (4D) scan slices and made viewable through, for example, a web application, and made manipulable through a user interface thereof.