Real-Time MR Imaging Using a Pre-Learned Spatial Subspace

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Generating on-the-fly MR images with sufficient temporal resolution for MR-based therapies and treatments is challenging due to the complexity involved, leading to slow image generation.

Innovation Solution

A method and system utilizing a pre-learned spatial subspace for real-time MR imaging, involving k-space data decomposition and transformation to rapidly construct dynamic images, enabling rapid image generation with high temporal resolution.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If conventional spatio-temporal decomposition methods are used for real-time MRI, then image generation can be performed, but the process is very slow and cannot achieve sufficient temporal resolution

Engineering Contradiction:
Improveimage generation speedVSAvoidtime delay in tracking therapies
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-computing and storing a spatial subspace basis during an initial training phase. This pre-learned spatial subspace is then reused during real-time imaging to rapidly construct dynamic images without repeating the full decomposition process, thereby achieving fast image generation with temporal resolution under 50 milliseconds.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If high temporal resolution is achieved through rapid image construction, then real-time tracking is enabled, but the complexity of the image generation process increases

Engineering Contradiction:
Improvetemporal resolutionVSAvoidcomplexity of image generation process
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the image generation process into two distinct phases: an offline training phase where the spatial subspace is pre-computed, and an online real-time phase where only simple linear combinations are performed. This segmentation allows high temporal resolution to be achieved during real-time imaging while the computational complexity is confined to the initial training phase, not affecting the real-time performance.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4146065B1Generating 3D dynamic images using pre-learned spatial subspace
Publication Date: 2025.12.24 CEDARS SINAI MEDICAL CENT
  • EP4146065B1 patent drawingFigure 1
  • EP4146065B1 patent drawingFigure 2
  • EP4146065B1 patent drawingFigure 3

AI summary

A method for performing real-time magnetic resonance (MR) imaging on a subject is disclosed. A prep pulse sequence is applied to the subject to obtain a high-quality special subspace, and a direct linear mapping from k-space training data to subspace coordinates. A live pulse sequence is then applied to the subject. During the live pulse sequence, real-time images are constructed using a fast matrix multiplication procedure on a single instance of the k-space training readout (e.g., a single k-space line or trajectory), which can be acquired at a high temporal rate.