Deep Learning Multiplex MRI Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Multiplex MRI image reconstruction methods face challenges in reconstructing parametric maps from subsampled data, which prolongs scan time and complicates the process.
Innovation Solution
A machine learning model, specifically a deep learning model, is trained to directly reconstruct parametric maps from subsampled Multiplex MRI data without first reconstructing echo images, utilizing different sampling masks and coil compression to enhance information recovery and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If subsampled data is acquired to speed up the acquisition process, then scan time is reduced, but reconstruction quality deteriorates
Solution Approach 1:
A deep learning model is introduced as an intermediary between the subsampled k-space data and the final parametric maps. The model learns to bridge the information gap created by subsampling, enabling accurate reconstruction without requiring full sampling. This intermediary system allows the use of subsampled data while maintaining reconstruction quality.
Solution Approach 2:
The patent transforms the reconstruction problem by changing the approach from traditional iterative reconstruction methods to a learned mapping approach. The deep learning model is trained to directly map subsampled data to parametric maps, fundamentally changing the reconstruction parameters and methodology to accommodate subsampled input while maintaining output quality.
2Ease of manufacture
If traditional two-step reconstruction method is used, then reconstruction process is systematic, but computational complexity increases
Solution Approach 1:
The patent merges the two separate reconstruction steps (echo image reconstruction and parametric map generation) into a single unified deep learning model. This single model performs both functions simultaneously, reducing computational overhead and simplifying the overall reconstruction process while maintaining systematic organization.
Solution Approach 2:
The deep learning model is designed with multi-functionality, capable of performing both echo image reconstruction and parametric map generation within a single framework. This universal approach eliminates the need for separate specialized algorithms for each step, reducing overall computational complexity while preserving the systematic nature of the process.
3Ease of operation
If multiple echo images are reconstructed separately, then each image can be processed independently, but information correlation is lost
Solution Approach 1:
The patent adds a new dimension to the processing by stacking multiple echo images along the channel dimension of the neural network input. This dimensional transformation allows the model to simultaneously process multiple echoes while preserving their correlations, effectively moving from independent 2D processing to correlated multi-dimensional processing.
Data Source
AI summary
In Multiplex MRI image reconstruction, a hardware processor acquires sub-sampled Multiplex MRI data and reconstructs parametric images from the sub-sampled Multiplex MRI data. A machine learning model or deep learning model uses the subsampled Multiplex MRI data as the input and parametric maps calculated from the fully sampled data, or reconstructed fully sample data, as the ground truth. The model learns to reconstruct the parametric maps directly from the subsampled Multiplex MRI data.