Personal Prior Information Constrained Low Dose CT Reconstruction Algorithms

Computed Tomography(CT)is a widely applied tool in clinical examination due to the high temporal and spatial resolution,fast imaging and noninvasive examination.However,excessive X-ray exposure arouses the public concern about unexpected damage to normal tissues.Many researches have revealed that excessive radiation is associated with cancers and other genetic diseases.Reducing the number of X-ray projections is supposed to lower the radiation dose.But reconstruction using the insufficient X-ray projection data would exhibit severe streak artifacts.Thus CT studies have aimed at decreasing radiation dose and yielding equal reconstruction quality.Recently many low dose CT reconstruction algorithms have been proposed based on Compressed Sensing(CS)theory and image noise removal.However,only taking use of the image sparsity is insufficient to restore image details.Multiple CT scans would be required for one patient in some cases,such as follow-up examination,perfusion imaging,and image guided therapy.These scans show similarity between each other.Full use of the prior information could be helpful to improve the image quality of the patient's low dose CT scan.This study proposed low dose CT reconstruction algorithms based on the similarity in prior images and the CS theorem:(1)Regional prior patch based few-view CT reconstruction method.This method proposes numerous prior image patches to represent structural details.In experiments,few-view CT scans are simulated from the Shepp-Logan model and clinical lung CT images.Root-mean-square error(RMSE)and peak signal to noise ratio(PSNR)are adopted to evaluate the reconstruction.The results indicate that by using patient's prior CT images,the low dose CT scan image could be improved with reserved structural details and less artifacts.(2)Personal historical lung CT scan based few-view CT ROI scan and reconstruction.This method aims at lung nodule follow-up and includes: 1.ROI centered limited X-ray beam regional CT scan;2.Regional prior patch based ROI reconstruction.ROI scanning is simulated from projecting lung CT images.Jaccard and Dice coefficients are applied to evaluate the nodule size and nodule shape,which are the key factors for nodule diagnosis.ROI reconstruction results show that the method is able to extremely lower the amount of X-rays,and reserve nodule details.