CT影像组学特征在肺泡蛋白沉积症中的诊断价值OA北大核心CSTPCD

Objective To investigate the diagnostic value of CT radiomics features in pulmonary alveolar proteinosis(PAP).Methods The general data and clinical characteristics of 24 patients with PAP in the Chinese PLA General Hospital from November 2008 to August 2022 were retrospectively collected and analyzed.Another 53 patients with other diffuse lung diseases except for PAP during the same period served as control group.The differences in the 10 conventional CT signs(semantic features)and 107 radiomics features between the two groups were compared.All patients were randomly divided into the training group(n=53)and the validation group(n=24)at a ratio of 7:3.CT semantic feature model,radiomics model and combined model to diagnose PAP were constructed in training group,and the diagnostic efficacy of models was compared using the receiver operating characteristic(ROC)curve in validation group.Decision curve analysis(DCA)was used to assess the value of models for practical clinical application.Radscore was calculated for the model with the highest diagnostic efficacy.Results A total of 24 patients with pathologically confirmed PAP were enrolled,with a male to female ratio of 3:1 and an average age of(44.6±15.2)years.The main clinical symptoms of patients with PAP included shortness of breath,cough,sputum and chest tightness.Compared with control group,the incidence of pleural effusion in PAP group was significantly lower(P＜0.05),while no significant differences were observed in other CT features(P＞0.05).The areas under the curve(AUC)of the semantic feature model for diagnosing PAP in training and validation group were 0.590 and 0.594,respectively,and in validation group,the accuracy,sensitivity,and specificity for diagnosis of PAP were 0.188,1.000,and 0.188,respectively.The AUCs of the radiomics model in training group and validation group were 0.845 and 0.867,respectively,and in validation group,the accuracy,sensitivity,and specificity were 0.641,0.938,and 0.703,respectively.The AUCs of the combined model in training group and validation group were 0.850 and 0.883,respectively,and in validation group,the accuracy,sensitivity,and specificity were 0.688,0.750,and 0.938,respectively.The AUCs of the radiomics model and the combined model were significantly greater than that of the semantic feature model,but there was no significant difference in the AUCs between the first two models.The decision curve analysis showed that both the radiomics model and the combined model had high application value for predicting PAP.Conclusion CT radiomics shows higher clinical value in the diagnosis of PAP compared with conventional CT features.