LDCT image super-resolution enhancement method and device based on residual convolutional neural network

Abstract

The invention discloses an LDCT image super-resolution enhancement method and device based on a residual convolutional neural network. According to the method, firstly, a network structure of an improved hybrid cascade task convolutional neural network (taking U-Net as an example) is designed, and then according to the designed network structure, a large number of LDCT low-resolution images and truth-value high-resolution images are provided for network training. The network training process is formed by performing feature extraction, error calculation and error back propagation on an improved hybrid cascade task U-Net, and back propagation is performed on an error value. The specified learning rate is 0.0001, the optimizer is ADAM, the learning rate adopts a stage decreasing strategy, and the loss between the super-resolution CT image and the true-value high-resolution image is continuously reduced. The method can be suitable for screening, detecting and analyzing three diseases (pulmonary nodules, chronic obstructive pulmonary diseases and coronary heart diseases) and the spine through one-time chest LDCT scanning in the medical field, and extra high-precision CT scanning does not need to be conducted on a certain local part of the chest with the higher spatial resolution requirement.

Description

technical field [0001] The invention relates to the field of medical LDCT image processing, in particular to a residual convolution neural network-based LDCT image super-resolution enhancement method and device. Background technique [0002] From 1970 to 1980, CRX (chest radiography) and sputum cytology were used as tools for lung cancer screening. The four randomized controlled trials failed to reduce lung cancer mortality. In the 1990s, CT came out, and LDCT (low-dose CT) could detect early lung cancer, but the mortality rate did not decrease. In the 2000s, low-dose CT screening could reduce lung cancer mortality by 20% compared with chest X-rays. After 2011, LDCT became popular worldwide. Artificial intelligence can be applied in various stages of accurate diagnosis and treatment of pulmonary nodules, including early screening of the population, intelligent early diagnosis, precise early treatment, full follow-up management, translational scientific research, etc. [00...

AI Technical Summary

Problems solved by technology

[0003] The main method currently used in the medical field is to use high-dose CT to screen, detect and analyze pulmonary nodules, chronic obstructive pulmonary disease, coronary heart disease and other diseases. The disadvantages of these methods are mainly slow imaging speed, high cost and high radiation Wait

Due to the low-dose CT used in LDCT, the imaging resolution is low. In particular, the existing artificial intelligence methods are only analyzed on low-resolution LDCT, which cannot meet the technical requirements of complex situations.

Images