Neural Network De-noising for Accelerated Wave-CAIPI MRI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional 3D magnetic resonance imaging (MRI) sequences face challenges with long scan times, leading to reduced patient throughput and increased susceptibility to patient motion, due to the intrinsic signal-to-noise ratio (SNR) loss in highly accelerated Wave-CAIPIRINHA scans.
Innovation Solution
The use of a neural network-based approach for de-noising highly accelerated Wave-CAIPIRINHA scans, combining fast Wave scans with image de-noising techniques to reduce scan time and improve SNR, particularly suitable for neuro-protocols like Wave MPRAGE, SPACE T2w, and SWI, and applicable even on low-end MRI systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional 3D MRI sequences are used, then high signal-to-noise ratio and isotropic resolution are achieved, but scan time becomes excessively long
Solution Approach 1:
The patent applies Wave-CAIPI encoding which modifies the k-space sampling parameters by introducing wave-like phase modulation patterns. This parameter change enables highly accelerated scanning (R=3×3 or higher) while maintaining acceptable signal-to-noise ratio through optimized coil sensitivity encoding and controlled aliasing patterns.
Solution Approach 2:
The patent segments the k-space sampling into multiple bands or shells with different acceleration factors. By applying different Wave-CAIPI encoding patterns to different segments, the system can achieve high overall acceleration while distributing the signal-to-noise ratio burden across multiple manageable segments rather than requiring uniform high acceleration throughout.
2Loss of time
If highly accelerated Wave-CAIPIRINHA scans are used, then scan time is reduced, but signal-to-noise ratio loss increases proportionally to sqrt(R)
Solution Approach 1:
The patent introduces a neural network as an intermediary processing step between the accelerated Wave-CAIPI acquisition and the final image reconstruction. This neural network mediator learns to predict and correct the sqrt(R) signal-to-noise ratio degradation, effectively decoupling the acceleration factor from the noise penalty through data-driven compensation.
Solution Approach 2:
The patent optimizes the Wave-CAIPI encoding parameters (wave frequency, amplitude, and phase patterns) specifically to minimize the g-factor penalty at high acceleration factors. By carefully tuning these parameters, the system achieves near-optimal g-factor performance (∼1) even at R=3×3 acceleration, thereby reducing the intrinsic sqrt(R) noise penalty.
3Productivity
If conventional parallel imaging methods are applied, then spatial encoding is achieved, but g-factor noise penalty and scan time reduction are limited
Solution Approach 1:
The patent employs dynamic Wave-CAIPI encoding where the phase modulation patterns are continuously varied across different k-space lines and partitions. This dynamic approach allows optimal exploitation of coil sensitivity variations at each sampling point, achieving superior g-factor performance (∼1) compared to static parallel imaging methods and enabling higher acceleration factors without proportional noise penalties.
Data Source
AI summary
Techniques are disclosed to leverage the use of neural networks or similar machine learning algorithms to de-noise highly accelerated Wave-CAIPIRINHA scans. The described techniques facilitate the generation of 3D sequences using a greatly reduced scan time, with the resulting images having a high spatial resolution and an improved SNR compared to conventional approaches.