SPECT imaging prediction model creation method and apparatus, and device and storage medium

An electronic device, comprising:

• a memory, configured to store a computer program;
• a processor, configured to execute the computer program to cause the electronic device to:
• obtain a training set comprising a plurality of scanning image groups, wherein each of the scanning image groups comprises a standard acquisition duration single-photon emission computed tomography (SPECT) image and a short acquisition duration SPECT image, and a-computed tomography (CT) image corresponding to a same subject being acquired;
• acquire, by a single-photon emission computed tomography device, a standard acquisition duration γ particle signal according to a standard acquisition duration;
• obtain a corresponding short acquisition duration γ particle signal by performing down acquisition on the standard acquisition duration γ particle signal, the standard acquisition duration γ particle signal and the short acquisition duration γ particle signal being under a same acquisition condition including a starting time of acquisition, the subject being acquired, a radioisotope drug dosage, and an acquisition angle;
• reconstruct the standard acquisition duration γ particle signal and the short acquisition duration γ particle signal by using a reconstruction algorithm, so as to obtain the standard acquisition duration SPECT image and the short acquisition duration SPECT image;
• perform network construction on the basis of a U2-Net network structure to obtain a network to be trained, wherein the network to be trained comprises a plurality of serially connected encoders and decoders, each level of the encoders and decoders consisting of a residual U-block (RSU), each RSU comprising 3-5 levels of serially connected sub-encoders and sub-decoders, the sub-encoders and sub-decoders comprising 3×3 convolutional layers, batch normalization layers, and rectified linear unit nonlinear layers connected by 2×2 pooling layers or up-sampling layers, with dilation convolution mixing receptive fields and skip connections introducing feature information at corresponding scales into an up-sampling process;
• and
• take the short acquisition duration SPECT image and the CT image in the training set as input-side training data, take the standard acquisition duration SPECT image in the training set as output-side training data, and train the network to be trained to obtain a SPECT imaging prediction model.

A non-transitory computer-readable storage medium, configured to store a computer program, wherein when the computer program is executed by a processor of an electronic device, the electronic device is caused to:

• obtain a training set comprising a plurality of scanning image groups, wherein each of the scanning image groups comprises a standard acquisition duration single-photon emission computed tomography (SPECT) image and a short acquisition duration SPECT image, and a computed tomography (CT) image corresponding to a same subject being acquired;
• acquire, by a single-photon emission computed tomography device, a standard acquisition duration γ particle signal according to a standard acquisition duration;
• obtain a corresponding short acquisition duration γ particle signal by performing down acquisition on the standard acquisition duration γ particle signal, the standard acquisition duration γ particle signal and the short acquisition duration γ particle signal being under a same acquisition condition including a starting time of acquisition, the subject being acquired, a radioisotope drug dosage, and an acquisition angle;
• reconstruct the standard acquisition duration γ particle signal and the short acquisition duration γ particle signal by using a reconstruction algorithm, so as to obtain the standard acquisition duration SPECT image and the short acquisition duration SPECT image;
• perform network construction on the basis of a U2-Net network structure to obtain a network to be trained, wherein the network to be trained comprises a plurality of serially connected encoders and decoders, each level of the encoders and decoders consisting of a residual U-block (RSU), each RSU comprising 3-5 levels of serially connected sub-encoders and sub-decoders, the sub-encoders and sub-decoders comprising 3×3 convolutional layers, batch normalization layers, and rectified linear unit nonlinear layers connected by 2×2 pooling layers or up-sampling layers, with dilation convolution mixing receptive fields and skip connections introducing feature information at corresponding scales into an up-sampling process;
• and
• take the short acquisition duration SPECT image and the CT image in the training set as input-side training data, take the standard acquisition duration SPECT image in the training set as output-side training data, and train the network to be trained to obtain a SPECT imaging prediction model.