Time-frequency transform and phase correction method of magnetic resonance signal based on residual network

Abstract

The invention discloses a time-frequency transform and phase correction method of a magnetic resonance signal based on a residual network, related to time-frequency transform methods of magnetic resonance signals. The time-frequency transform and phase correction method of the magnetic resonance signal based on the residual network comprises the following steps: constructing a mathematical model of a free-sensing attenuation time-domain signal having zero-order phase imbalance and first-order phase imbalance simultaneously and an ideal absorption spectrum according to the characteristics of aFID signal, that is, constructing a mathematical model of the spectrum real part without phase imbalance, and generating a simulation signal by means of the mathematical model to construct training set data and test set data; constructing a residual network model and setting relevant training parameters; performing network training; and performing network testing. The free-sensing attenuation signal in the time domain is automatically transformed into the ideal absorption spectrum by using the residual neural network in deep learning, that is, the spectrum real part with the phase imbalance does not exist. In the process, the automatic conversion from the time domain to the frequency domain and the automatic correction of the phase imbalance are completed, so that the ideal absorption spectrum can be obtained.

Description

technical field [0001] The invention relates to a method for time-frequency transformation of magnetic resonance signals, in particular to a method for time-frequency transformation and phase correction of magnetic resonance signals based on a residual network. 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. However, in practical applications, there are often various types of phase imbalance in magnetic resonance spectroscopy, so the data preprocessing of magnetic resonance spectroscopy is an indispensable part of the spectrum analysis process. Phase offset generally includes zero-order offset and first-order offset. Zero-order offset refers to the phase difference between the reference phase and the free induction decay signal (Free Induction Decay, FID) received by the instrument [1] . The fi...

AI Technical Summary

Problems solved by technology

This is not only very time-consuming, but also the quality of the corrected spectrum depends on the operator's experience, so fast and automatic phase correction is very necessary, that is, changing the manual / visual phase correction process into automatic phase correction

Some traditional methods that have been proposed are often not sufficiently accurate and robust for automatic phase correction

Images