Novel image reconstruction system for nuclear medicine through training the neural network for improving the spatial resolution and image quality simultaneously based on structural image of phantoms

摘要

The present invention relates to a new nuclear medicine image reconstruction system through neural network learning for simultaneous improvement of spatial resolution and image quality based on structural images for a phantom, comprising: a phantom unit for generating a phantom including various spatial frequencies and brightness of pixels; an imaging unit that takes a nuclear medicine image or a structural image based on the phantom; a label generating unit that fuses the two images taken by the imaging unit to generate a nuclear medicine image that does not contain blurring necessary for learning as a label image; and a learning unit that sets the sinogram data of the nuclear medicine image and the label image as input data and output label, respectively, and derives a correct answer image by removing blurring and noise included in the input data through neural network learning. According to the present invention as described above, by measuring the structural phantom while repeatedly changing the position, resolution correction is possible without measuring the PSF or LSF for all spaces, and daily quality control (DQC) performed in hospitals ), it is possible to add measurements using a structural phantom to the process, so that it can be applied more flexibly in the field through the continuous update of the correction algorithm according to the change of the system that changes over time.