Differential Diagnosis And Prognosis Of Pneumonia-type Invasive Mucinous Adenocarcinoma By Combination Of Clinical And CT-based Radiomics Signature

Part Ⅰ Differential diagnosis of pneumonia-type invasive mucinous adenocarcinoma and pneumonia by combination of clinical and CT-based radiomics signatureObjectivePneumonia-type invasive mucinous adenocarcinoma(IMA)is a variant subtype of lung adenocarcinoma with special morphological characteristics.Due to its lack of specificity in clinical and imaging manifestations,it is often misdiagnosed as infectious pneumonia,which leads to delayed treatment and poor prognosis.The purpose of this study was to develop and validate a predictive model combining clinical and CT-based radiomics signatures to differentiating pneumonia-type IMA from pneumonia.MethodsThis retrospective study included 314 patients from six hospitals between January 2017 and January 2022 who underwent noncontrast chest CT showing consolidation and were diagnosed with pneumonia-type IMA(n=106)or pneumonia(n=208).Patients from three hospitals formed a training set(n=195)and a validation set(n=50)in an 8:2 ratio,and patients from the other three hospitals formed the external test set(n=69).A clinical model for predicting pneumonia-type IMA was built using clinical characteristics(including sex,age,cough,sputum,fever,smoking history,family history of lung cancer,elevated white blood cell count levels,and elevated C-reactive protein levels)that were significant independent predictors of this diagnosis.Radiomics features were extracted from CT images by placing ROIs on areas of consolidation,and the best features were selected using the least absolute shrinkage and selection operator(LASSO)method,then a radiomics score(Rad-score)of pneumonia-type IMA was constructed.A nomogram for predicting pneumonia-type IMA was constructed that combined features in the clinical model and the Rad-score.Two cardiothoracic radiologists independently reviewed CT images in the external test set to diagnosis pneumonia-type IMA or pneumonia.Diagnosis performance was compared among models and radiologists.Decision curve analysis(DCA)was performed.ResultsThe clinical model was constructed by two clinical factors,fever and family history of lung cancer.The area under the curve(AUC)of the clinical model was 0.73(95%CI:0.68-0.78)in the training set,0.71(95%CI:0.59-0.83)in the validation set,and 0.71(95%CI:0.61-0.80)in the external test set.After LASSO regression analysis,15 imaging features were used to construct the radiomics model.In the training set,validation set,and external test set,the rad-score of pneumonia-IMA patients were significantly higher than those of pneumonia patients(0.50±1.88 vs-1.97±1.44,P<0.001;-0.25±1.22 vs-1.75±1.54,P=0.001;-0.63±2.05 vs-2.68±1.34,P<0.001).The AUC of the radiomics model in the training set,validation set and external test set were 0.87(95%CI:0.82-0.91),0.81(95%CI:0.69-0.92)and 0.81(95%CI:0.70-0.90),respectively.The AUCs of nomogram combining clinical factors and radiomics signatures in the training set,validation set,and external test sets were 0.93(95%CI:0.89-0.95),0.82(95%CI:0.72-0.91),and 0.85(95%CI:0.76-0.92),respectively.DeLong test showed that the diagnostic efficacy of nomogram was higher than that of clinical model in all three study groups(all P<0.05).DCA showed higher overall net benefit from the nomogram than from the clinical model.In addition,AUCs of two cardiothoracic radiologists in the external test set were 0.70(95%CI:0.58-0.81)and 0.67(95%CI:0.55-0.78),respectively,and the diagnostic efficacy of nomogram was higher than that of two radiologists(both P<0.05).ConclusionThe nomogram combined with clinical and CT-based radiomics features could distinguish pneumonia-type IMA from pneumonia,and its diagnostic accuracy is significantly higher than that of clinical model and two radiologists,which could help guide appropriate clinical management decisions.Part Ⅱ Survival prediction of pneumonia-type invasive mucinous adenocarcinoma by combination of clinical and CT-based radiomics signatureObjectivePneumonia-type IMA is prone to intracavitary spread leading to intrapulmonary metastasis and is thought to be associated with poorer prognosis.Besides,the biological behavior of pneumonia-type IMA is highly variable,with significant differences in response to treatment and prognosis among individuals.The objective of this study was to establish a prediction model for survival outcomes of pneumonia-type IMA combing clinical factors and CT-based radiomics signature,and to evaluate the predictive efficacy of the model.MethodsThis retrospective study recruited 129 patients of pneumonia-type IMA from six hospitals between January 2017 and July 2022.The training set(n=90)and external test set(n=39)were randomly divided according to a ratio of 7:3.Survival analyses included distant metastasis-free survival(DMFS)and overall survival(OS).Follow-up was conducted every 3-6 months for the first year and annually thereafter.DMFS was defined as the time between diagnosis of pneumonia-type IMA and discovery of the first distant metastatic tumor(such as the brain,liver,bone,abdominal cavity,or other distant parts of the body)or death.OS was defined as the time from diagnosis of pneumonia-type IMA to death from any cause.For patients without distant metastasis or death,follow-up time in the survival analysis was defined as the last follow-up.Clinical factors and CT findings were analyzed by univariate and multivariate Cox regression analysis to establish clinical models of DMFS and OS.LASSO Cox regression was used to select the optimal radiomics signatures to develop the radiomics models.Integrating clinical-imaging factors and radiomics signatures,the nomogram was developed.The performance of models was determined by C-index,decision curve,and calibration curve.DMFS and OS probabilities were estimated using the Kaplan-Meier and compared by log-rank test.ResultsDMFS and OS were similar for both the training set and external test set(all P>0.05,log-rank test),the median follow-up time was 35 months(range 5-82 months)and 32 months(range 9 to 102 months),respectively.The multiple Cox regression analysis showed that maximum tumor diameter(P=0.028),interlobar fissure expansion(P=0.031),interstitial change(P=0.020),pleural effusion(P=0.012)and treatment method(P=0.037)were independent risk factors for DMFS.And interleaf fissure expansion(P=0.003),interstitial change(P=0.006),pleural effsion(P=0.018)and treatment method(P=0.016)were associated with OS.The C-index values of clinical model for predicting DMFS were 0.792(95%CI:0.731-0.853)and 0.678(95%CI:0.561-0.795)in the training and external test set,respectively;and for OS were 0.816(95%CI:0.750-0.882)and 0.726(95%CI:0.612-0.839),respectively.With a median of 0.011 for DMFS and 0.062 for OS,patients were divided into low-risk and high-risk groups.There was significant difference between the two risk categories in OS(P<0.001 for both training set and external test set),but no significant difference in DMFS for the external test set(P=0.15).The radiomics model for predicting DMFS and OS were developed based on 18 and 12 radiomics signatures,respectively.The C-indexes for DMFS were 0.778(95%CI:0.717-0.839)and 0.704(95%CI:0.585-0.823)in the training set and external test set,respectively;and for OS were 0.783(95%CI:0.710-0.855)and 0.716 0.716(95%CI:0.584-0.848),respectively.With the median of 0.484 for DMFS and 0.533 for OS,there were statistically significant between low-risk and high-risk groups(all P<0.05).Nomogram showed a better predictive performance,with a C-index of 0.830(95%CI:0.773-0.887)and 0.721(95%CI:0.608-0.834)for DMFS,and 0.851(95%CI:0.793-0.909)and 0.788(95%CI:0.700-0.876)for OS in the training set and external test set.With the median of-0.065 for DMFS and-0.192 for OS,there were statistically significant between low-risk and high-risk groups(all P<0.05).The DCA demonstrated that nomogram provided greater net benefit,and the calibration curve showed that it was in good agreement with the observed outcomes.ConclusionThe nomogram,integrating clinical factors and radiomics signatures,performed a good predictive performance for prediction distant metastasis or death in pneumonia-type IMA,which could help guide follow-up strategies and treatment,and hopefully prolong survival.