Value Of CT Quantitative Study Of Functional Small Airway And Lung Vessel In Early Diagnosis Of COPD

Part 1Quantitative CT Evaluation of Functional small airway in Early Diagnosis of COPD【Objective】To explore the quantitative CT evaluation of functional small airway in early diagnosis of COPD,and to explore the predictive performance for high risk COPD with artificial intelligence,in order to provide more sensitive image markers for early diagnosis of COPD and lay a foundation for the early intelligent warning diagnosis of COPD.【Methods and Materials】From August 2018 to December 2019,615 subjects with complete questionnaire survey,lung function test（PFT）and paired inspiratory and expiratory chest CT scanning were enrolled retrospectively.According to FEV1/FVC and FEV1%,the subjects were divided into normal group（FEV1/FVC>70%,n=367）,high risk group（FEV1/FVC>70%,80%<FEV1%<95%,n=194）and COPD group（FEV1/FVC<70%,n=54,including GOLD I 14,GOLD II 32,GOLD III+IV 8）.The CT quantitative parameter of small airway is lung volume,parameter response mapping（PRM）,including volume（cc）and volume percentage（%）of emphysema,functional-small airways disease（f SAD）,normal lung tissue（Normal）,and unknown classified lung tissue（Uncategorized）on the level of whole lung,right and left lung,and each lobe,respectively.There are 72parameters at the level of whole lung,left and right lung and 5 lobes.The following statistical analysis was analyzed:（1）Differences in 4 basic clinical characteristics（age,gender height and weight）,13 PFT parameters and 72 CT quantitative parameters were analyzed among the 3 groups.（2）The correlation between PRM parameters and pulmonary function parameters was analyzed.（3）Deep learning was performed to establish a model predicting PFT based on PRM,in order to realize that one-stop routine CT scanning can evaluate the functional small airway and predict PFT index.（4）The redefinition of high-risk COPD was analyzed with deep learning based on PRM parameters to select the optimal FEV1%threshold.【Results】（1）Univariate analysis showed that there was no significant difference in age among the three groups（P>0.05）,but there was significant difference in height and weight among the three groups（P=0.00）.There were statistical differences in 12 parameters of PFT and 68 parameters of small airway（P<0.05）among the 3 groups:the mean values of PRMV^Emph,PRMV^Emph%,PRMV^{f SAD},PRMV^{f SAD}%in whole lung,right lung,left lung and each lobe in normal group,high-risk group and COPD group increased in turn.PRMV^Normal%of the whole lung,right lung,left lung,right lower lobe and left upper lobe decreased in the normal group,high-risk group and COPD group in turn.The lung volume（LV）,the mean values of PRMV^Normal,PRMV^Normal%,PRMV^{Uncategorized}and PRMV^{Uncategorized}%of the whole lung,right lung and left lung in the normal group were greater than those in the high-risk group.（2）The correlation between PRM parameters and PFT parameters was mostly low.Only PRMV^Normaland PRMV^{Uncategorized}were moderately positively correlated with FVC（r=0.3-0.6,P<0.05）.The lung volumes、PRMV^Normal、PRMV^{Uncategorized}of whole lung,left lung,right lung and each lobe were moderately positively correlated with FEV1.0（r=0.3-0.6,P<0.05）.It can be seen that PRMV^Normalor/and PRMV^{Uncategorized}are moderately correlated,PFT parameters were weakly or extremely weakly correlated with PRMV^{f SAD},PRMV^{f SAD}%,PRMV^Emph,PRMV^Emph%.（3）The performance of PRM in predicting PFT parameters:for predicting FEV1/FVC,R²was 0.864 and 0.749 in training and validation set,respectively;for predicting FEV1%,R²was 0.888 and 0.792 in training and validation set,respectively.The predictive performance of this model was good enough.The combined predictive model composed of PRM parameters and four clinical indexes by confusion matrix showed that the sensitivity,specificity and accuracy for differentiating normal group from high-risk group were 0.85,0.90 and 0.88,respectively;and the sensitivity,specificity and accuracy for differentiating non COPD group from COPD group were 0.89,1 and 0.99,respectively.While the accuracy for subclassification of COPD（GOLD I,GOLDII and GOLD III+IV）was only 0.44.（4）For the redefining the high-risk COPD,the deep learning based on PRM parameters showed that the consistency between PRM parameters and PFT for distinguishing normal from high-risk COPD was good enough and AUC was 0.83 when FEV1%threshold was 0.72;which outweighed the current common FEV1%threshold of0.95 with AUC 0.64.【Conclusion】The CT quantitative parameter PRM of small airway could distinguish normal population from high-risk COPD and COPD population on the level of whole lung and each lobe.The regression predictive model based on PRM parameter showed good performance differentiating normal group from high risk group,and differentiating non-COPD group from COPD group.Therefore,one-stop CT scan could evaluate the functional small airway and PFT simultaneously.Deep learning based on PRM parameters to redefine high-risk COPD was feasible,and the optimal setting of FEV1%threshold would lay the foundation for the redefinition of COPD diagnosis.Part 2CT Quantitative Evaluation of Lung Vessel in Early Diagnosis of COPD【Objective】To evaluate the CT quantitative parameters of lung vessels in the early diagnosis of COPD and analyze the focal distribution features of small vessel remodeling,helping the selection of early imaging warning markers of COPD.【Materials and methods】1112 patients from 3 thoracic diseases screening in our hospital from August 2018 to December 2019 were included and divided into three groups according to FEV1/FVC and FEV1%（same as part I）,normal group（n = 657）,high risk group（n = 379）and COPD group（n = 76）.CT quantitative parameters of small vessels included the number of pulmonary vessels（Num Vessels（ea））,cross-sectional area less than 5mm2（Cross sectional area < 5,CSA < 5）,number of small vessels（Num Vessels Under 5mm2（ea））,average vessel diameter（Diameter Mean（mm））,average vessel area(Area Mean（mm2）,vessel surface density（Surface Mean（HU））,emphysema area（Surface LAA（mm2））（CT value-950HU）,vessel area（Vessel Area（mm2））,total vessel surface area（Total Surface Area（mm2））;and all the above parameters of the whole lung at the level of 6mm,9mm,12 mm,15mm,18 mm,21mm and 24 mm to the costal pleura were analyzed,56 parameters totally.The following statistical analysis was performed:（1）Differences were analyzed in 3 basic data characteristics（age,sex and BMI）,13 pulmonary function parameters and 56 quantitative parameters of small vessels among the three groups.（2）The correlation was performed between the parameters of small vessels and the corresponding parameters of pulmonary function.（3）The R language rstatix tool was used to evaluate the gradient trend of CT quantitative parameters of small vessels among different populations（normal,high risk,GOLD I,II,III+IV）from 6mm to 24 mm to the costal pleura.【Results】1 Univariate analysis showed that there were significant differences in sex,BMI,smoking history,family history of emphysema,concomitant diseases,clinical symptoms（cough,asthma and shortness of breath）among the three groups.Moreover there was significant difference in BMI between the normal group and the high-risk group（P <0.033*）.The average vessel area（Whole Lung₆mm,18²4mm）and vessel area（Whole Lung₁2,18,21mm）increased in the normal group,high-risk group and COPD group in turn,while the average surface density（Whole Lung₆mm²4mm）and vessel area（Whole Lung₁5mm）in the normal group were less than those in the high-risk group and greater than those in the COPD group.The mean vessel number（Whole Lung₆,9mm）and CSA < 5（Whole Lung₆,15mm）decreased in normal group,high risk group and COPD group in turn,and the emphysema surface area（Whole Lung₆²4mm）,total surface area（Whole Lung₆²4mm）,CSA < 5（Whole Lung₉/12/18²4mm）and vessel area（Whole Lung₂4mm）in normal group were greater than those in high-risk group and less than those in COPD group.（2）Most of the parameters of small vessels showed weak or no correlation with pulmonary function parameters（r < 0.2,P < 0.05）.Taking the whole lung level as an example,only FVC was moderately positively correlated with the vessel number in the whole lung（15²4mm from the pleura）and the vessel number with CSA < 5（r=0.307⁰.629,P < 0.05）.The total surface area showed significant positive correlation with FEV1/FVC（r=0.601⁰.629,p<0.05）.FEV1.0 was positively and moderately correlated with the vessel number in the whole lung（12²4 mm from the pleura）,vessel number with CSA < 5 and the total surface area（r=0.429⁰.556,p<0.05）.（3）According to the trend of small vessels from the different distance to the costal pleura,the trend of 8 parameters including Area Mean（mm2）,Vessel Area（mm2）,Num Vessels（ea）,CSA < 5,Diameter Mean（mm）,Surface Mean（HU）,Surface LAA（mm2）,Total Surface Area（mm2）were consistent among the 3 groups from 6²4mm to the costal pleura: the Area Mean（mm2）,Diameter Mean（mm）and Surface Mean（HU）showed a gradual increasing trend;Num Vessels（ea）,CSA<5,Surface LAA（mm2）and Total Surface Area（mm2）decreased gradually;the Vessel Area（mm2）increased at 6⁹mm,then decreased gradually at 9²4mm.The parameters of high-risk group were greater than those of normal group including Area Mean（mm2）,Diameter Mean（mm）,Surface Mean（HU）and Vessel Area（mm2）.The following parameters of normal group were greater than high risk group,Num Vessels（ea）,CSA < 5,Surface LAA（mm2）and Total Surface Area（mm2）.Area Mean（mm2）（about 9mm-15mm）and Vessel Area（mm2）（about6mm¹2mm）in non-COPD group were greater than those in COPD group.The parameters of COPD group were greater than those of non-COPD group including Surface LAA（mm2）,Total Surface Area（mm2）,Num Vessels（ea）（between 9²4mm）and CSA < 5（between 9²4mm）.Compared with 9mm and 15 mm to pleura,there was a turning point in the distinction between non-COPD group and COPD group,that is,at9mm¹5mm,non-COPD group was greater than COPD group,and within the range of less than 9mm and greater than 15 mm,non-COPD group was less than COPD group.The Vessel Area（mm2）,Num Vessels（ea）and CSA < 5 in 9mm to 24 mm distance to pleura in COPD group was greater than those in non-COPD group,but there were no significant differences between high-risk group and normal group.【Conclusion】Eight CT quantitative parameters of lung vessels,including Area Mean（mm2）,Vessel Area（mm2）,Num Vessels（ea）,CSA < 5,Diameter Mean（mm）,Surface Mean（HU）,Surface LAA（mm2）,Total Surface Area（mm2）,could distinguish normal group,high risk group from COPD group.There was a gradient distribution of small vessel quantitative parameters at different distances to costal pleura,which would lay a foundation and theoretical basis for the quantitative evaluation of small vessels at the level of lung lobe and segment.