| Objective:The judgment of the negative conversion time of pulmonary tuberculosis is the key to determine the therapeutic effect of tuberculosis.Based on the computer-aided diagnosis of pulmonary infection,this study quantitatively analyzed the chest CT of invasive pulmonary tuberculosis patients with delayed negative conversion of tuberculosis bacteria,studied the diagnostic value of imaging quantitative indexes in patients with pulmonary tuberculosis,and discussed the related predictive factors of negative conversion delay combined with clinical indexes,so as to improve the degree of attention of doctors and patients to the disease in the early stage.To provide a basis for clinical rationalization of treatment.This study is divided into two parts:Part one:To verify the accuracy of pulmonary infection assistant diagnosis system based on quantitative CT in the detection of pulmonary lesions in patients with invasive pulmonary tuberculosis,and to explore the correlation between visual score,CT quantitative index and clinical laboratory examination index.Part two:Based on the quantitative CT assistant diagnosis system for pulmonary infection,the characteristics of CT quantitative indexes in patients with tuberculosis not negative after 2 months of treatment were analyzed,and the diagnostic efficacy of CT quantitative indexes in delayed pulmonary tuberculosis was discussed.To further construct the prediction model of pulmonary tuberculosis with delayed negative conversion,and to find the predictive indexes for early identification of patients with delayed negative conversion combined with clinical data.Part one1.Materials and methodsA total of 103 patients with invasive pulmonary tuberculosis were included in the affiliated Hospital of Yan’an University and the second people’s Hospital of Yan’an City.The original data of CT scanning were imported into"digital lung"artificial intelligence computer aided diagnosis and analysis platform for quantitative analysis of pulmonary infection in"Dicom"format.After the completion of treatment,the quantitative indexes of pulmonary lesions,such as lesion quality(LM),lesion volume(Le V),mean lesion density(MLe D)and lesion proportion(Le V%)were derived.The correlation between artificial vision score and lesion quality,lesion proportion and lesion volume obtained by AI assistant diagnosis system was tested by Pearson and Spearman,and the correlation between laboratory index and artificial vision score and CT quantitative parameters was further analyzed.2.ResultsThe score of lesion severity evaluated by physicians was highly correlated with lesion volume(r=0.834,P<0.01),lesion proportion(r=0.804,P<0.01)and lesion quality(r=0.801,P<0.01).The proportion of lesions was moderately positively correlated with NEU%(r=0.504,P<0.01)and ESR(r=0.519,P<0.01),slightly positively correlated with PLT(r=0.304,P<0.01),moderately negatively correlated with PAB(r=-0.580,P<0.01),and slightly negatively correlated with LYM%(r=-0.494,P<0.01)and A/G(r=-0.360,P<0.01).The lesion quality was moderately positively correlated with NEU%(r=0.514,P<0.01)and ESR(r=0.511,P<0.01),slightly positively correlated with PLT(r=0.305,P<0.01),moderately negatively correlated with LYM%(r=-0.514,P<0.01)and PAB(r=-0.559,P<0.01),and slightly negatively correlated with ALB(r=-0.314,P<0.01)and A/G(r=-0.361,P<0.01).3.Brief summaryIn this study,the lesion results obtained by the auxiliary diagnosis system of pulmonary infection and the visual score of the imaging doctor were analyzed,and it was found that the lesion parameters quantitatively obtained by the software had a strong correlation with the physician’s visual evaluation of the focus severity score,and a weak and moderate correlation with some clinical indexes,which proved the feasibility of AI in the quantitative analysis of pulmonary tuberculosis lesions and could be further applied to the clinical evaluation of pulmonary tuberculosis lesions.Part two1.Materials and methods313 patients with pulmonary tuberculosis diagnosed and examined by bronchoscopy in Yan’an second people’s Hospital were collected retrospectively.157 patients without CT data and 33 patients with negative tuberculosis were excluded.40 patients were not reexamined on time after the end of the intensive period of treatment,and 83 patients were included finally,including 50 patients with tuberculosis negative and 33 patients without negative at the end of 2 months of treatment.A patient who does not turn negative at the end of the treatment enhancement period(at the end of 2 months)is defined as a delay in negative conversion.The statistics were analyzed by SPSS software and R language.Chi-square test was used to compare the counting data between the two groups.In the comparison of measurement data between groups,the T-test of two independent samples is used when the data are in accordance with the normal distribution,and the Mann-Whitney rank sum test is used when the data is not in line with the normal distribution.ROC curve analysis of the difference between the two groups of quantitative parameters in the diagnosis of patients with delayed negative conversion.First,single factor logistic analysis was used to screen the independent risk factors of mycobacterium tuberculosis negative conversion delay,and then these factors were analyzed by multivariate logistic regression analysis to further select the predictive factors of tuberculosis negative delayed conversion and establish a multivariate logistic regression model.To establish model 1 which includes only clinical indicators and model 2 which includes both clinical and quantitative indicators.Then the AUC values in the ROC analysis of the two models were used to evaluate the discrimination ability of the model.The improvement of model 2(clinical-quantitative imaging index regression model)compared with model 1(clinical regression model)was judged by NRI and IDI analysis.The difference was statistically significant.The difference was statistically significant(P<0.05).2.ResultsAt the end of 2 months after treatment,the proportion of males(?~2=4.849,P<0.05)in the non-negative group was higher than that in the negative group.There were no significant differences in age(Z=-0.619,P>0.05),nature of residence(?~2=0.103,P>0.05),smoking(?~2=3.733,P>0.05),BMI(Z=-1.466,P>0.05)and the time from symptom to visit(Z=-0.621,P>0.05).There was no significant difference in the incidence of pleural effusion between the two groups(?~2=1.912,P>0.05).The incidence of cavity(?~2=17.694,P<0.05)and more than 3 lobes involved in the chest(?~2=12.662,P<0.05)in the non-negative group was significantly higher than that in the negative group.In the laboratory indexes,compared with the negative conversion group,the white blood cell count(Z=-5.304,P<0.05),neutrophil count(Z=-4.964,P<0.05),monocyte count(t=-4.397,P<0.05)and globulin(t=-2.328,P<0.05)in the non-negative conversion group were significantly higher than those in the negative conversion group,while the lymphocyte percentage(t=2.191,P<0.05)and the ratio of A/G ratio(Z=-2.27,P<0.05)were lower in the negative conversion group.There was no significant difference in platelet(Z=-1.95,P>0.05),red blood cell count(t=-1.315,P>0.05)and hemoglobin(t=-1.127,P>0.05)between the two groups(P>0.05).The quantitative CT index lesion volume(Le V)(Z=-5.229,P<0.05),lesion proportion(Le V%)(Z=-4.513,P<0.05)and lesion quality(LM)(Z=-5.08,P<0.05)in the non-negative group were significantly higher than those in the negative group.There was no significant difference in mean lesion density(MLe D)between the two groups(Z=-0.735,P>0.05).CT quantitative indexes Le V,Le V%and LM can be used to distinguish between negative patients and non-negative patients after 2 months of treatment.When Le V>69.74ml,Le V%>1.495%and LM>48.84g,the areas under the curve of patients classified as non-negative patients were 0.841,0.794 and 0.831,respectively.The results of multivariate stepwise logistic regression analysis showed that NEU,cavity and Le V%were independent predictors of negative delay.Using these predictive factors to establish two models to predict the negative conversion delay of tuberculosis,and then using ROC curve analysis,it was found that the AUC of model 2 was higher than that of model 1,and the AUC values of the two models were 0.870 and 0.852respectively.NRI and IDI analysis between model 2(clinical-quantitative imaging index regression model)and model 1(clinical regression model)showed that the predictive efficiency of model 2 was significantly better than that of model 1(NRI=1.599,P<0.001;IDI=0.465,P<0.001).3.Brief summaryIn this study,the pulmonary infection assistant diagnosis system based on AI was used to analyze the pulmonary lesions,and the lesion volume,lesion proportion and lesion quality were obtained.it was found that the patients with delayed negative conversion had large lesion volume,wide range and large lesion quality.when the whole lung lesion volume reached 69.74ml,the lesion proportion reached 1.495%and the lesion mass reached 48.84g,the patients with negative delay could be better identified.A prediction model of quantitative parameters combined with clinical characteristics was established to predict the delayed conversion of Mycobacterium tuberculosis.Three predictive factors of neutrophil absolute value,cavity and lesion ratio were included in the model.The final prediction model has good prediction efficiency and can help respiratory physicians identify patients with delayed negative conversion at the initial stage of the disease and further optimize the diagnosis and treatment plan.ConclusionsIn the correlation analysis between the automatic analysis results of the assistant diagnosis system of pulmonary infection and the severity score of artificial vision evaluation,the quantitative lesion parameters were highly correlated with the physician’s score of lesion severity,the physician’s score of lesion severity was slightly correlated with some clinical inflammatory indexes,and the quantitative parameters were slightly correlated with some inflammatory indexes,which proved that quantitative CT was accurate and practical in the evaluation of pulmonary tuberculosis.In the study of pulmonary tuberculosis patients with and without tuberculosis negative conversion after 2 months treatment,it was found that the patients with delayed tuberculosis negative conversion had a wide range of pulmonary lesions and large lesion mass.When the whole lung lesion volume reached 69.74ml,the lesion proportion reached 1.495%,and the lesion mass reached 48.84g,the patients with delayed negative conversion could be better identified.A prediction model of quantitative parameters combined with clinical characteristics was established to predict the delayed conversion of Mycobacterium tuberculosis.Three predictive factors of neutrophil absolute value,cavity and lesion ratio were included in the model.The predictive model has good predictive efficiency and can help respiratory physicians identify patients with delayed negative conversion at the initial stage of the disease and further rationalize treatment. |
PDF Full Text Request