Magnetic particle imaging (mpi) reconstruction method based on recnet model

An MPI reconstruction system based on a RecNet model, wherein the system comprises a signal obtaining module and an image reconstruction module, wherein

• the signal obtaining module is configured to obtain a 1D MPI signal on which imaging reconstruction is to be performed, and take the 1D MPI signal as an input signal;
• and
• the image reconstruction module is configured to input the input signal and a velocity signal of an FFP corresponding to the input signal into a trained magnetic particle reconstruction model RecNet for image reconstruction to obtain a 2D MPI image, wherein
• a method for obtaining the velocity signal of the FFP comprises:
• calculating a position of the FFP in an x direction:
• xffp=HDx/Gx
• wherein x ffp represents the position of the FFP in the x direction, HDx represents magnetic field intensity of a driving field, and Gx represents a change amount of unit-distance magnetic field intensity of a selection field in the x direction;
• and
• obtaining a derivative of the position of the FFP in the x direction with respect to time to obtain the velocity signal of the FFP;
• the magnetic particle reconstruction model RecNet comprises a domain conversion network and an improved UNet network;
• the domain conversion network is constructed based on a 1D convolutional layer, a fully connected layer, a concatenation layer, and a spatial feature extraction layer that are sequentially connected;
• the 1D convolutional layer is configured to perform 1D convolution processing on the 1D MPI signal and the velocity signal of the FFP separately;
• the fully connected layer is configured to convert an obtained 1D MPI signal after the 1D convolution processing into a 2D signal that is taken as a first signal, and convert an obtained velocity signal of the FFP after the 1D convolution processing into a 2D signal that is taken as a second signal;
• and the concatenation layer is configured to concatenate the first signal and the second signal, and take a concatenated signal as a third signal;
• the spatial feature extraction layer is constructed based on a plurality of sequential connected convolutional layers, a batch normalization layer, and a nonlinear activation function, and the spatial feature extraction layer is configured to perform feature extraction on the third signal;
• and
• the improved UNet network is constructed based on a UNet network and a plurality of 1×1 convolutional layers, and the 1×1 convolutional layers are connected to an output layer of the UNet network;
• and the improved UNet network is configured to further extract a feature extracted by the spatial feature extraction layer, and perform image reconstruction to obtain the 2D MPI image.