positron emission computed tomography (PET) reconstruction method based on an automatic encoder network

Abstract

The invention provides a positron emission computed tomography (PET) reconstruction method based on an automatic encoder network. The method comprises the following steps: step A, on the basis of a positron emission computed tomography (PET) image, establishing a de-noising auto-encoder (DAE) network model, and obtaining prior information of the image by using the trained de-noising auto-encoder (DAE); Step B, combining a de-noising auto-encoder (DAE) network with image prior information with a traditional PET iterative reconstruction method; the two images are alternately iterated to obtain areconstructed image; According to the invention, a de-noising auto-encoder (DAE) network is integrated on the basis of a positron emission computed tomography (PET) reconstruction method; A DAE network is used for unsupervised learning of priori information of a PET image, then the DAE network with the image priori information and a traditional PET iterative reconstruction method are combined, the DAE network and the traditional PET iterative reconstruction method are alternately iterated to obtain a reconstructed image, and a good reconstruction effect is achieved.

Description

technical field [0001] The invention relates to the technical field of medical image processing, and is mainly used in the fields of computerized tomography (CT), positron emission computed tomography (PET), denoising, restoration and reconstruction of medical noise-containing images, and specifically a A reconstruction method for positron emission tomography (PET) based on an autoencoder network. Background technique [0002] Positron emission computed tomography (PET) is currently the only new imaging technology that can display biomolecular metabolism, receptors, and neurotransmitter activities in vivo. It is an important tool for tumor research and clinical diagnosis and treatment. PET reconstruction is a method to reconstruct clinician-acceptable functional images from projection sinusoidal data with low technical rate and noise influence. Due to detector blurring effect, positron range, and photon non-collinearity, existing PET reconstruction images have problems such...

AI Technical Summary

Problems solved by technology

Since the MLEM algorithm is an ill-posed problem to solve the inverse problem of the emission distribution from the measurement data, once the iteration reaches a certain point, the iterative expectation maximization algorithm for the maximum likelihood solution will lead to increased noise

While the MAP algorithm eliminates divergence at higher iterations, traditional smoothing priors or total variation previously lead to oversmoothing or artifacts in the reconstructed image

[0006] Existing positron emission tomography (PET) reconstruction methods based on autoencoder networks suffer from large noise, excessive smoothing, and artifacts

Images