Intelligent quantitative analysis method for magnetic resonance spectrum based on deep learning

Abstract

The invention discloses a magnetic resonance spectrum intelligent quantitative analysis method based on deep learning, and relates to a magnetic resonance spectrum quantitative analysis method. The magnetic resonance spectrum intelligent quantitative analysis method based on deep learning can achieve quantitative analysis of target metabolism signals and has the advantages of being high in quantification speed, high in quantification precision and intelligent and integrated. Constructing simulation data by using a magnetic resonance spectrum signal model, wherein the simulation data comprises a magnetic resonance spectrum signal under a simulated non-ideal imaging condition, a corresponding metabolite component signal and a background signal spectrum, and the magnetic resonance spectrum signal, the corresponding metabolite component signal and the background signal spectrum serve as a simulation training set; designing a deep learning spectrum quantization network and a loss function; training the network by using the obtained simulation training set to obtain the network under the optimal parameters; inputting the living magnetic resonance spectrum into the network under the optimal parameters to obtain a background signal, a target metabolite component signal and a signal parameter of a metabolite component under a non-ideal imaging condition; and calculating the concentration of the target metabolite by using the signal parameters obtained by network prediction.

Description

technical field [0001] The present invention relates to a quantitative analysis method of magnetic resonance spectrum, in particular to an intelligent quantitative analysis method of magnetic resonance spectrum based on deep learning. Background technique [0002] Magnetic resonance spectroscopy (Magnetic Resonance Spectroscopy, MRS) is widely used in non-invasive biological metabolic detection, in which proton magnetic resonance spectroscopy can non-invasively quantify the concentration of metabolites in the brain, and is used to analyze a variety of diseases, such as epilepsy, Multiple sclerosis, stroke, cancer and metabolic diseases, etc. Due to the signal characteristics and experimental conditions of MRS, the challenges of its research and analysis lie in: (1) There are many kinds of metabolites in the body, which lead to serious overlapping of spectral peaks and it is difficult to separate them; (2) The acquisition of MRS signals is susceptible to noise interference, a...

AI Technical Summary

Problems solved by technology

However, research on quantitative methods for MRS deep learning is not extensive

Images