Deep learning sparse angle CT reconstruction method based on frequency domain and image domain degradation perception

Abstract

The invention discloses a deep learning sparse angle CT (Computed Tomography) reconstruction method based on frequency domain and image domain degradation perception, which belongs to the technical field of image processing, and comprises the following steps: firstly, constructing a frequency domain model, designing a frequency domain attention module, displaying and learning different characteristics of sparse angle CTs with different degradation levels in the frequency domain, and outputting weighted frequency characteristics; sending to a frequency domain reconstruction module to generate a frequency domain reconstruction image; and secondly, constructing an image domain model, designing an image domain attention module, learning edge pixel reconstruction features of sparse angle CTs of different degradation levels in an image domain by using a frequency domain reconstruction image, outputting an image domain attention prediction image, and finally sending the image domain attention prediction image into the image domain reconstruction module to output a final reconstruction result. By combining a CT data set containing multiple degradation levels and supervised training, the method can overcome the defects that an existing deep learning method facing a single degradation level is poor in generalization and cannot be expanded, the overall precision of reconstruction is effectively improved, noise and artifacts are restrained, and meanwhile detailed texture features are reserved.

Description

technical field [0001] The invention belongs to the technical field of image processing, and more specifically, relates to a deep learning sparse-angle CT reconstruction method based on frequency domain and image domain degradation perception. Background technique [0002] CT reconstruction is the process of reconstructing the projection data obtained by CT scanning into CT images through algorithms. Over the past fifty years, CT images have been widely used in clinical diagnosis, nondestructive testing, and biological research due to their high resolution and high sensitivity. With the continuous development of medical CT technology, people put forward requirements for faster, safer and higher precision CT technology. However, high doses of radiation can cause headaches and, in severe cases, even cancer and leukemia. Additionally, long scan times and high scan frequencies further increase the hazard. Sparse-angle CT and limited-angle CT reduce the number of measurements ...

AI Technical Summary

Problems solved by technology

However, current deep learning reconstruction methods often apply supervised learning on data with a single degree of degradation (such as sparse projection data from 120 angles), so they cannot be used on data with other degrees of degradation (such as sparse projection data from 60 angles). Get good refactoring performance

A straightforward way to solve this problem is to train a set of parameters for each degradation level, but it is a great challenge to deploy in practice due to the growth of training computation and parameter storage

In addition, as the level of degradation increases, the training cost and parameter storage increase linearly, which is not scalable in practical applications

Another way to alleviate this problem is to construct a training dataset by mixing data of multiple degradation levels, but this will cause the model to predict a compromised reconstruction result, and there is still room for improvement in performance

Images