Machine Learning-based Age-related Lung Changes And Lung Function Imaging Studies

Part I Low Dose CT scanning of Non-Calcified Pulmonary Nodules by Artificial Intelligence:A large-scale Multi-center Analysis in ChinaPurpose Pulmonary nodules are increasingly detected with the aid of low-dose computed tomography(LDCT)and artificial intelligence(AI).This study aimed to investigate the prevalence of non-calcified pulmonary nodules(NCPNs)in the Chinese population.Materials and methods A total of 94,892 participants with chest LDCT were included in multi-center between Oct 2017 and Oct 2018.NCPNs were firstly recognized by AI and subsequently confirmed by two radiologists and two respiratory physicians.The prevalence of NCPNs was investigated as well as other detailed information regarding the number of pulmonary nodules,their location,and characteristics,as interpreted by two radiologists.Results The overall incidence of NCPNs was 59.3%(38,051/64,168,95%CI 0.588-0.598).The incidence of NCPNs in male was 61.7%(21,193/34,349,95%CI 0.609-0.623),numerically higher than that in female 56.5%(16,848/29,819,95%CI 0.551-0.57.6).Gender and age are risk factors for PNs.In 188,744 detected NCPNs,77.9%were solid nodules(SNs),3.2%were part-solid nodules(PSNs)and 18.9%were ground glass nodules(GGNs);27.1%were in right upper lobe(RUL),12.8%in right middle lobe(RML),21.1%in right lower lobe(RLL),19.6%in left upper lobe(LUL)and 19.4%in left lower lobe(LLL);13.2%were 2-4 mm,63.7%were 5-8 mm and 23.1%were 9-30 mm.ConclusionsWith the help of AI,the big data analysis shows a high prevalence of NCPNs in LDCT screening.The right upper lobe is the most common location of lung nodule distribution.Most of the detected NCPNs are solid nodules,and their size is between 5-8mm.Attention should be paid to the problem of over detecting of pulmonary nodules by artificial intelligencePart ? A novel prediction model for the development of small airway dysfunction evaluation with radiomicsPurpose Small airway dysfunction is a common but easily neglected respiratory abnormality.Little is known about its risk factors,prevalence and prognostic factors.The aim of this study was to explore whether small airway dysfunction could be early detected on CT images with radiomics analysis,so as to predict the progress of small airway dysfunction.Method The data of patients who underwent pulmonary function examination in our medical center from July 2017 to August 2020 were retrospectively collected.The data of their previous pulmonary function examination and PACS imaging data were retrospectively analyzed,and 272 patients were included.The patients were divided into small airway dysfunction group and non-small airway dysfunction group.The diagnostic criteria for small airway dysfunction were on the basis of at least two of the following three indicators of lung function being less than 65%of predicted:maximal mid-expiratory flow(MMEF),forced expiratory flow(FEF)50%,and FEF 75%.The patients were randomly divided into training group(n=190)and validation group(n=82)with a ratio of 7:3.Lung kit software was used for automatic delineation of region of interest(ROI)on chest CT images.The most valuable imaging features were selected by lasso regression,and the radiomics score was established for risk assessment.Univariate and multivariate regression analysis were used to determine the independent risk factors and establish the final prediction model.C-index,calibration curve and DCA curve were used to evaluate the accuracy of the model and tested by the validation group.Results Multivariate logistic regression analysis showed that age,rad-score,smoking and history of asthma were significant predictors of small airway dysfunction(P<0.05).The area under the curve(AUC)predicted by imaging score was 0.842[95%confidence interval(CI)0.784-0.900](training cohort)and 0.856(95%CI0.765-0.948)(validation cohort),respectively.The area under the curve for predicting small airway dysfunction using the nomogram model was 0.910(95%CI 0.860-0.946)(training cohort)and 0.8933(95%CI 0.856-0.977)(validation cohort),respectively.Analysis shows that nomogram has potential generalization ability.DCA curve shows that the prediction model has high clinical application value.Conclusion Radiomics showed potential value for early assessment of small airway dysfunction.Our model may provide respiratory physician with useful information and might serve as reliable reference for clinical treatment strategies of small airway dysfunction.