Decoupling method for magnetic resonance signals based on deep learning
A magnetic resonance signal and deep learning technology, which is applied in neural learning methods, magnetic resonance measurement, and measurement using nuclear magnetic resonance spectrum, etc., can solve the difficulties of unreliable chemical shift identification, sensitivity loss, suppression or removal of homonuclear coupling, etc. question
Active Publication Date: 2020-11-10
XIAMEN UNIV
View PDF9 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
In the experiment, the coupling between nuclei causes multiple splits in the signal, which increases the complexity of the spectrum and reduces the resolution and sensitivity.
The extra dimension introduced in 2D NMR experiments such as TOCSY and NOESY can reduce the effect of signal overlap, but in many applications the spectral peak overlap is still serious, resulting in unreliable chemical shift assignment and affecting the analysis of molecular structure
However, suppressing or removing homonuclear coupling is difficult, and practical decoupling methods have not been available until recently [4-7]
The pure chemical shift spectrum obtained by this method contains only one signal for each chemically different site, which can remove the splitting caused by ordinary coupling, but there is a great loss in sensitivity [8]
In addition, it is also accompanied by problems such as false peaks, high experimental difficulty and long experimental time.
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0041] The following embodiments will further illustrate the present invention in conjunction with the accompanying drawings.
[0042] In this specific embodiment, J-coupling removal is performed on the magnetic resonance spectrum, and noise filtering is performed to obtain an ideal absorption spectrum. The input spectrum length is N 1 =1×4001, the length of the obtained output spectrum is N 2 =1×4001.
[0043] figure 1 Given is the network model for decoupling and denoising. exist figure 1 In, the length size is N 1 The magnetic resonance spectrum of =1×4001 is used as the input of the network, and the network is divided into two parts, the left part is the contraction path, and the right part is the expansion path, and most of the two paths are left-right symmetrical structures. Each step of the contraction path is composed of two layers of 1×3 convolutions (filling convolution), each layer of convolution is followed by a nonlinear unit ReLU, and each step is followed ...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
A method for decoupling magnetic resonance signals based on deep learning, involving magnetic resonance signals. According to the characteristics of the magnetic resonance spectrum signal, construct the mathematical model of the network input data and the network label, that is, the mathematical model of the real part of the spectrum without the J-coupling phenomenon, generate the simulation signal from the mathematical model, construct the training set data and the test set data; build the network model, set the relevant training parameters; input the training set data into the network model to train the network, adjust the network parameters until the loss function decreases to converge and stabilize, and obtain a functional network model; input the test set data signal into In the functionalized network model after training, the ideal absorption spectrum obtained by network decoupling is obtained, and compared with the label to verify the performance of the network. Realize decoupling and denoising functions, realize end-to-end functions, do not need other auxiliary methods such as preprocessing of spectral signals, and use neural networks to truly realize decoupling and denoising functions.
Description
technical field [0001] The invention relates to a J-coupling method for magnetic resonance signals, in particular to a J-coupling method for magnetic resonance signals based on deep learning. Background technique [0002] Magnetic resonance (NMR) is widely used in the fields of chemometric analysis and biotechnology, and a wealth of molecular structure information can be obtained from magnetic resonance spectroscopy. Resolution and sensitivity are critical for extracting detailed molecular structural information using NMR spectroscopy, and since the advent of NMR spectroscopy [1-3] , people are constantly seeking new ways to improve the resolution and sensitivity of homonuclear correlation experiments. However, in the experiment, the coupling between nuclei causes multiple splits in the signal, which increases the complexity of the spectrum and reduces the resolution and sensitivity. The extra dimension introduced in 2D NMR experiments such as TOCSY and NOESY can reduce th...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G01R33/46G06N3/04G06N3/08
CPCG01R33/4625G06N3/084G06N3/045
Inventor 索斐杨钰蔡聪波陈忠
Owner XIAMEN UNIV