Research On Image Enhancement Of Low-dose Pulmonary CT And Its Auxiliary Diagnostic Application

Lung cancer is currently one of the malignant tumors with the highest morbidity and mortality in the world,early detection and surgical removal will significantly improve the prognosis of it,however,most of the early lung cancers have no obvious symptoms but appear most in advanced stages.But traditional CT scans is harmful to patients and medical staff due to their large radiation doses,and low-dose lung CT caused widespread concern because it greatly reduces the radiation dose,but it also reduces the image quality and clarity.This article designs image enhancement and its auxiliary diagnostic applications based on low-dose lung CT images and assists doctors in the diagnosis and classification of lung nodules.In the past research,traditional spatial and frequency-domain enhancement algorithms have many limitations.This paper make a new theoretical perspective,analyzes the variational method in the theoretical rationality of low-dose lung CT image enhancement and enhance them based on image decomposition,and design a reasonable decomposition model,this algorithm can be greatly improved in terms of contrast,sharpness,detail texture,and visual effects to achieve better observation results.Then,we analyzes the shortcomings of traditional machine learning and deep learning algorithms in lung nodule detection and classification,designs an algorithm on low-dose CT lung nodules location with target segmentation.This algorithm based on traditional U-Net model,increase the residual unit and the three-dimensional network structure named RU3D-Net,which greatly increases the detection accuracy of micro-small nodules,and the false positive rate reaches 92.73%.In the lung nodule classification task,we made three-dimensional structure into the Res-net structure create a Res3 D network,so that the final classification accuracy can reach 99%.Finally we fusion the image enhancement algorithm,the lung nodule detection algorithm and the lung nodule classification algorithm into auxiliary diagnosis application on low-dose lung CT images.The practical degree of each part is tested separately,which proves the value of this application system.