Assignment of Mr Fingerprints on the Basis of Neural Networks

Simple SummaryContent extracted from patent full text and abstract with AI.

This invention uses a trained neural network to directly determine magnetic resonance (MR) parameters, such as T1 and T2 relaxation times, from 'MR fingerprints'—unique signal sequences obtained from MRI scans. Instead of relying on conventional methods that match measured fingerprints to large, memory-intensive databases of simulated signals, the neural network rapidly predicts the tissue parameters from the MR fingerprint input, offering speed and efficiency improvements in MR image analysis.

Use CasesContent extracted from patent full text and abstract with AI.

• Medical imaging, especially MRI for rapid tissue characterization and disease diagnosis.
• Development of faster and more efficient MRI machines in hospitals and clinics.
• Creation of compact medical software for MR parameter mapping in portable or point-of-care devices.
• Research studies requiring high-throughput MR data analysis, such as neuroscience or oncology.
• Integration into MRI workflow automation for improved throughput and reduced operator workload.

BenefitsContent extracted from patent full text and abstract with AI.

• Significantly faster determination of MR parameters compared to traditional dictionary-matching methods.
• Reduces data storage requirements by eliminating the need for large simulation databases (dictionaries).
• Enables real-time or near real-time analysis and mapping of tissue characteristics.
• Can improve diagnostic accuracy and consistency by leveraging machine learning to handle noise and artifacts.
• Allows MR parameter estimation for a large number of voxels simultaneously, supporting high-resolution mapping.
• Provides flexibility to handle different MR acquisition protocols and parameter combinations.
• Potential for use in a wide range of MRI systems, including those with limited computing resources.

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

In a method for determining magnetic resonance (MR) parameters, an MR fingerprint of a voxel is acquired by execution of a pulse sequence, the MR fingerprint is provided as an input into the input layer of a trained neural network, and at least one MR parameter relating to the MR fingerprint is provided at the output layer of the neural network.