Font Size: a A A

Image Quality Of Down Sample CTP With A New Reconstruction Algorithm:CNN Based On DRL

Posted on:2021-01-12Degree:MasterType:Thesis
Country:ChinaCandidate:J N TianFull Text:PDF
GTID:2404330626459116Subject:Imaging and nuclear medicine
Abstract/Summary:PDF Full Text Request
Purpose:The radiation dose of the CTP was reduced by lowering the sampling frequency,from the original 30-pass frequency sampling to 15-pass.Then a innovatve image reconstruction model-DRL based on Convolutional Neural Networks(CNN)image reconstruction model,The algorithm is used to restore the of the 15-pass down sample CTP image to 30-pass sampling frequency.The aim of this study was to evaluate the image quality when using 30-pass and 15-pass CTP.Methods:We analyzed a total of 236 patients with suspected acute ischemic stroke in the first hospital of jilin university from june 2018 to january2019.All patients were scanned with 30-pass sampling frequeny CTP by brilliance ict 256(philips),The 15-pass sampling frequency CTP image is obtained by removing the odd number image by engineering software.Then the new image reconstruction model DRL was used to restore the15-pass sampling frequency CTP image.The restored 15-pass and original 30-pass images were input into the PMA-ASIST(japan)post-processing software to generate two sets of perfusion color maps and obtain different perfusion parameters: CBV?CBF?MTT and TTP.All imaging findings were interpreted and evaluated by two experienced neuroimaging physicians who did not know the patient's information.PSNR and SSIM indexes were used to evaluate the quality of 15-pass and 30-pass orignal perfusion images.To evaluate the quality of perfusion color maps in 15 pass and 30 pass groups,the CBV?CBF?MTT and TTP color maps of the two groups were evaluated and scored respectively.the ischemic region and the infarct core were further evaluated,and the ASPECTS scores of the two groups of image cbv and cbf color maps were calculated respectively.All data were statistically analyzed using SPSS 22.0 statistical software(SPSS,IBM,West Grove,PA,USA).Quantitative data expressed as meanąsd.Qualitative data are expressed as quantity and percentage.Use t test(for normally distributed data)or Mann-Whitney u test for quantitative data.P value<0.05 is consided statistically significant.Results:A total of 40 patients met the criteria for inclusion in the study.Original CTP image quality was similar for 30-pass versus 15-pass(PSNRavg=51.73,SSIMavg=0.9946,PSNRstd=1.4666,SSIMstd=0.0018).There is no difference in the quality of perfusion parameters of 30-passand 15-pass image.Corresponding Ischemic core lesions were slightly bigger at 15-pass versus 30-pass(P <0.05),whereas corresponding ischemic region showed excellent agreement.Both groups of images can meet the needs of clinical diagnosis.Conclusions:Reconstruction CTP images are typically still diagnostic.However,with regard to perfusion quantitation and image quality,our study indicates that origin frequency image is preferable to 15-pass.We recommend down-sampled CTP images combined DRL model as an option to reduce radiation dose in patients.
Keywords/Search Tags:CTP, radiation dose, Convolutional Neural Networks(CNN), image quality, reconstruction algorithm
PDF Full Text Request
Related items