Accelerated MR Thermometry Using Neural Network Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current MRI-guided focused-ultrasound treatments face inefficiencies due to the slow imaging rate of magnetic resonance (MR) imaging, which can lead to inadequate tracking of target changes during thermal ablation procedures, resulting in potential damage to healthy tissues or inefficiencies in treatment processes.
Innovation Solution
The implementation of a machine learning algorithm, such as a neural network, to estimate missing k-space data from incomplete data sets, allowing for accelerated MR thermometry by reconstructing images based on prior knowledge and iteratively updating information, thereby reducing the need for full k-space data acquisition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full k-space data acquisition is performed, then image reconstruction accuracy is improved, but imaging rate is reduced
Solution Approach 1:
The patent applies partial action by acquiring only a subset of k-space data (undersampling) rather than the complete data set. This allows the imaging rate to be accelerated while using machine learning algorithms to reconstruct accurate thermal maps from the incomplete data, effectively achieving good results with less than 100% of the required data points
Solution Approach 2:
The patent uses machine learning models trained on fully sampled k-space data to create accurate copies or predictions of what the complete thermal map would look like. The neural network learns the relationship between undersampled and fully sampled data, enabling it to generate accurate thermal maps from accelerated acquisitions
2Productivity
If MR imaging rate is increased to enable real-time monitoring, then treatment efficiency is improved, but image quality may deteriorate due to insufficient data
Solution Approach 1:
The patent introduces machine learning algorithms as an intermediary between the undersampled k-space data and the final thermal map reconstruction. This intermediary process fills in the missing information and corrects artifacts, enabling both high imaging rates and accurate thermal maps to coexist
Solution Approach 2:
The patent replaces the traditional mechanical/physical constraint of requiring full k-space data acquisition with a computational approach. Instead of physically acquiring all data points, machine learning computations are used to infer the missing information, substituting computational power for physical measurement time
3Loss of information
If conventional MR imaging is used for monitoring, then comprehensive thermal information is obtained, but the slow imaging rate causes delays in detecting target changes
Solution Approach 1:
The patent performs preliminary training of machine learning models using fully sampled k-space data before actual treatment monitoring. This preliminary action creates a trained model that can quickly reconstruct thermal maps from accelerated data during treatment, eliminating detection delays while maintaining information quality
Solution Approach 2:
The patent enables continuous real-time monitoring during treatment by using accelerated imaging with machine learning reconstruction. The system maintains continuous thermal map updates without the interruptions or delays that would occur with conventional slower imaging, ensuring uninterrupted treatment monitoring
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables real-time monitoring and accurate reconstruction of thermal maps during ultrasound procedures, improving the efficiency and precision of thermal ablation treatments by increasing the MR imaging rate and reducing the risk of damaging surrounding healthy tissues.
Implementation Method 1
magnetic resonance (MR) imaging rate
Implementation Method 2
Ultrasonic energy may be focused to a zone having a cross-section of only a few millimeters... During wave propagation through the tissue, a portion of the ultrasound energy is absorbed, leading to increased temperature
Implementation Method 3
The implementation of a machine learning algorithm, such as a neural network, to estimate missing k-space data from incomplete data sets
Data Source
AI summary
Systems and methods provide accelerated MR thermometry utilizing prior knowledge about the images to be reconstructed from incomplete k-space data, thereby facilitating accurate reconstruction. In various embodiments, missing data is computationally estimated using a machine learning algorithm such as a neural network, and an image is generated based on iteratively updated estimated missing information.