Four-dimensional cone beam CT reconstruction image enhancement algorithm based on N-net and CycN-net network structures

Abstract

The invention discloses a four-dimensional cone beam CT (Computed Tomography) reconstructed image enhancement algorithm based on N-net and CycN-net network structures. The method is designed on the basis of fully mining inherent characteristics of four-dimensional cone beam CT reconstructed images, the deep neural network CycN-net takes analysis reconstructed images of five continuous motion phases and corresponding prior images as network input, and takes prior knowledge in four-dimensional cone beam CT image sequences and space-time correlation among the image sequences into consideration; the extracted convolutional feature maps are stitched and fused to serve image restoration. The deep neural network CycN-net in the invention has excellent performance for recovering the four-dimensional cone beam CT image sequence with high temporal-spatial resolution, and can recover and maintain detail structure information in the reconstructed image to the greatest extent while effectively inhibiting serious bar artifacts.

Description

technical field [0001] The invention belongs to the technical field of X-ray imaging, in particular to a four-dimensional cone-beam CT reconstruction image enhancement algorithm based on N-net and CycN-net network structures. Background technique [0002] Radiation therapy is one of the main means of treating malignant tumors (commonly known as cancer). Cone Beam Computed Tomography (CBCT) is widely used in the field of radiation therapy, especially in Image Guided Radiation Therapy (IGRT). The main function of cone beam CT is to position the patient and locate the treatment target according to the reconstruction results, and at the same time provide real-time update of patient information for the radiotherapy process. Generally, the detectors of cone-beam CT in this field have high spatial resolution, which is very convenient for positioning the head, but it faces great challenges for lung imaging. The imaging time of cone beam CT is 1 minute to 4 minutes per week. During...

AI Technical Summary

Problems solved by technology

Although U-net has been successfully applied in many natural image restoration tasks, there are still limitations when facing 4D-CBCT image restoration tasks

A large number of experiments have shown that the detailed information in the 4DCBCT image predicted by the U-net network is largely lost, and the wrong artifact structure may also be introduced

Images