The Clinical Application Of Iterative Model Reconstruction With Pulmonary Ground Glass Nodules In Low Dose CT Through Artificial Intelligence

Part Ⅰ Experimental study of iterative model reconstruction with pulmonary ground glass nodules in low dose CTObjective Explore the best low dose scanning parameters of IMR algorithm for pulmonary GGN through simple chest model experiment,so as to prepare for the next clinical verification.Methods A simple chest model was made with fresh pig lung and acrylic plate.The scanning scheme is designed with tube voltage of 120 KV,100KV,80 KV and tube current of 40-150 m As.Each time the tube voltage is unchanged,the tube current is increased by 10 mas.A total of 36 scanning schemes are designed,and all of them are post processed by IMR algorithm.The image subjective score,the image noise value and radiation dose were measured.All the results were recorded and imported into excel and made into a scatter plot with smooth line for analysis.Results The image score value increases with the increase of tube current;the image noise value decreases with the increase of tube current;the radiation dose increases with the increase of tube current.Combined with subjective image score,noise value and radiation dose,the best low dose parameters were determined as tube voltage: 100 KV,tube current: 100 m As.Conclusion In the chest model test,the best scanning parameters of low dose CT combined with IMR algorithm for pulmonary GGN: tube voltage 100 KV,tube current 100 m As.This parameter can greatly reduce the radiation dose(58.42%),at the same time,it can display the detailed signs of lung GGN to the greatest extent and ensure the image quality.Whether the best low dose scanning parameters combined with IMR algorithm has value in the clinical application of chest CT remains to be further studied.Part Ⅱ Clinical study of iterative model reconstruction with pulmonary ground glass nodules in low dose CTObjective According to the low dose scanning parameters obtained from the previous experiment,combined with the iterative model reconstruction algorithm,the diagnostic ability of pulmonary ground glass nodules was studied,and the feasibility of clinical application was verified.Methods 3462 patients were scanned with low dose(100KV,100 m As)of chest CT.Finally 103 patients with lung GGN were enrolled in the study,the low dose group was recorded as the experimental group and the conventional dose group was recorded as the control group.The radiation dose,average CT value,noise value,SNR,subjective score,tumor lung interface,lobular sign,burr sign,nodule nature,pleural depression sign,tumor microvascular sign,air bronchogram sign and diagnosis accuracy rate were compared.The average CT value,noise value,SNR and subjective score of the two groups were tested by T-test;the sign statistics were expressed by frequency,and Mc Nemar test was used for the comparison between the two groups.Kappa was used to analyze the consistency of the scores between the observers.The difference was statistically significant(P < 0.05).Results The average CT value,noise value and SNR of the experimental group were difference with those of the control group,the difference was statistically significant(P ﹤0.05);There was no significant difference in the third branch of pulmonary artery、Bronchial branches of the dorsal segment of the lower lobe of the left lung、Subpleural parenchyma and Soft tissue of mediastinal window between the experimental group and the control group(P = 0.330、0.910、0.740、0.652);Theconsistency of subjective image scores between the experimental group and the control group was good(kappa value = 0.8271);The number of p GGN,m GGN typeⅠ,m GGN typeⅡ,microvascular sign,pleural indentation sign and air bronchus sign in the experimental group was the same as that in the control group;the clarity of leaf sign,prickle sign and tumor lung interface in the experimental group was slightly lower than that in the control group,and the difference was not statistically significant(P = 1.000、0.500、0.500);The diagnostic accuracy of benign and malignant nodules in the experimental group and the control group was 90.30%.The CDTIvol,dose length product(DLP)and effective dose(ED)of the experimental group were lower than those of the control group,the difference was statistically significant(P﹤0.05).Conclusion Low dose CT combined with IMR algorithm have high clinical application value in detection and diagnosis of lung GGN,when the radiation dose was reduced by 50.72% without affecting the detailed signs of lung GGN.Part Ⅲ The study of iterative model reconstruction with pulmonary ground glass nodules in low dose CT through Artificial intelligenceObjective Study on the diagnostic efficiency of low dose CT with iterative model reconstruction for pulmonary ground glass nodules through Artificial Intelligence.Methods 150 patients with pulmonary nodules were retrospectively analyzed.IMR algorithm is recorded as experimental group,i Dose algorithm is recorded as control group.AI was used to analysis and comparison of sensitivity,detection accuracy and prediction accuracy of GGN with different properties of two groups of images.The counting data was expressed by frequency.Mc Nemar test was used for comparison between the two groups,with P < 0.05 as the difference.Results The total sensitivity and sensitivity of pGGN in experimental group were higher than that of the control group,and the difference was statistically significant(P﹤0.05);The accuracy of m GGN type I,the sensitivity of m GGN type II,the total accuracy and the accuracy of pggn in the experimental group were lower than those in the control group,and the difference was not statistically significant(P=0.322、0.314、0.920、0.692);The accuracy of malignant nodule prediction in experimental group was higher than that of control group,and the difference was statistically significant(p=0.012);The prediction accuracy of microinvasive adenocarcinoma and adenocarcinoma in situ in the experimental group was higher than that in the control group,and the difference was not statistically significant(p=0.219、0.063).Conclusion In the case of low dose combined with AI,IMR algorithm has higher sensitivity and diagnostic accuracy than i Dose algorithm for detection of lung GGN,and has a high clinical application value.