| Lung cancer is the leading cause of cancer-related deaths worldwide,and the popularity of low-dose chest CT has revolutionized the global lung cancer screening landscape in recent decades.Especially in China,more and more asymptomatic ground-glass nodules(GGNs)are found in the lungs,and the incidence is also high in young women who do not smoke.There is a new trend in the risk factors of lung cancer,and the epidemiological has changed.Therefore,the differential diagnosis of GGN is increasingly challenging.Studies have reported that GGNs are more likely to be malignant than solid pulmonary nodules(34%vs.7%)and are mainly early adenocarcinomas,but can also be caused by benign lesions such as granulomas,interstitial pneumonia,and fungal infections.Existing diagnostic methods such as needle biopsy,fiberoptic bronchoscopy and high-resolution CT(HRCT)have insufficient diagnostic performance.Positron emission tomography/computed tomography(PET/CT)imaging has unique advantages in the diagnosis,efficacy evaluation and prognosis judgment of various tumors,but the monitoring ability of GGN is still controversial.Therefore,we conducted the following research to explore the diagnostic value of 18F-FDG PET/CT for GGN,and to construct a dual-mode non-invasive imaging diagnostic model to provide an important theoretical basis for improving the accurate diagnosis and treatment of early-stage lung adenocarcinoma.Part Ⅰ Association between 18F-FDG PET/CT-based SUV index and malignant status of ground-glass nodulesObjective:As an accurate non-invasive examination,18F-FDG PET/CT has high clinical application value for the diagnosis,staging and efficacy evaluation of solid pulmonary nodules,but there is a high false negative rate and false positive rate for the diagnosis of GGN type lung cancer,and traditional criteria for distinguishing benign from malignant is obviously not applicable to GGN.The relationship between the FDG metabolic parameters of GGN and its benign and malignant is the key to the differential diagnosis,and there are few related reports.Therefore,this study intends to explore the correlation between the standardized uptake value(SUV)index(SUVmax/SUVmean of liver)of GGN in 18FFDG PET/CT imaging and its malignant status.Methods:A total of 166 patients with GGN who underwent 18F-FDG PET/CT examination in our hospital from January 2012 to October 2019 were retrospectively analyzed,including 53 males and 113 females.A total of 192 GGNs were surgically removed,22 benign GGNs and 170 malignant GGNs were confirmed by pathology.The clinical and imaging data of all patients were collected,and GGNs were divided into three groups according to the SUV index,namely Tertile 1(0.14-0.54),Tertile 2(0.55-1.17)and Tertile 3(1.19-6.78),64 GGNs per group.Logistic regression was used to analyze the correlation between SUV index and malignant status of GGN,and generalized additive model was used to test the nonlinear relationship between them.Results:After adjusting for the potential confounding factors,we found that the malignancy risk of GGN significantly decreased as the SUV index increased(OR:0.245;95%CI:0.119-0.504;P<0.001).The average probability of malignant GGN was 89.1%(95%CI:53.1-98.3),80.5%(95%CI:36.7-96.7),and 34.3%(95%CI:9.5-72.2)for Tertile 1 to Tertile 3.And the increasing trend of SUV index was significantly correlated with the reduction of malignant risk(OR:0.099;95%CI:0.025-0.394;P=0.001),especially between Tertile 3 versus Tertile 1(OR:0.064;95%CI:0.012-0.356;P=0.002).Curve fitting showed that the SUV index was linearly negatively correlated with the malignant risk of GGN,and the correlation between the two groups was consistent in different age groups,GGN number,type,shape,whether accompanied by vacuolar sign and different consolidation-to-tumor ratio(CTR)groups,and there was no significant interaction(P=0.187-1.000).Conclusion:SUV index was an independent correlative factor of GGN malignancy risk.The higher the SUV index,the lower the malignant probability of GGN,and the two have an approximate linear negative correlation.Understanding the relationship between the them will help to improve the accuracy of 18F-FDG PET/CT in the differential diagnosis of benign and malignant GGN.Part Ⅱ Establishment and validation of a predictive model for the malignancy probability of ground-glass nodules based on 18F-FDG PET/CT imaging parameters and clinical featuresObjective:Noninvasive imaging examination is the main method to distinguish between benign and malignant ground glass nodules(GGNs)before surgery.Due to the different morphology of GGN and various imaging manifestations,single CT morphological features and quantitative parameters have limited diagnostic value for GGN.The construction of multi-factor prediction model is expected to improve the diagnostic performance of GGN.However,there is no well-recognized prediction model dedicated to distinguishing benign and malignant GGN at home and abroad.Therefore,this study intends to construct a multi-factor prediction model based on 18F-FDG PET/CT,in order to improve the diagnostic efficiency of distinguishing benign and malignant GGN.Methods:A retrospective analysis of 170 patients with GGN who underwent PET/CT examination in our hospital from November 2011 to December 2019,including 56 males and 114 females.The clinical and imaging data of all patients were collected,and the nodules were randomly divided into derivation set and validation set according to 1:1.For the derivation set,we used multivariate logistic regression to develop a malignancy probability prediction model of GGN.A receiver operating characteristic(ROC)curve was used to evaluate the diagnostic efficacy of the model,and the data in the validation set were used to verify the prediction model.Results:Among the 170 patients,a total of 197 GGNs were diagnosed by postoperative pathology or clinical follow-up,including 27 benign GGNs and 170 malignant GGNs.Logistic regression analysis showed that five indicators,including gender,nodule location,margin,pleural indentation and SUV index,were independent factors for predicting the malignant probability of GGN.The formula of the prediction model was as follows:Logit(P)=2.37643-1.79479 ×(female=0,male=1)-3.75834 ×(peripheral=0,center=1)+1.96792 ×(smooth=0,lobulated=1)+2.13431 ×(pleural indentation)-0.81624 × SUV index,where P is the predicted probability.The area under the curve(AUC)of the model was 0.875 in the derivation set,with a sensitivity of 0.702 and a specificity of 0.923.The positive likelihood ratio was 9.131,and the negative likelihood ratio was 0.322.In the validation set,the AUC of the model was 0.874,which was not significantly different from the derivation set(P=0.989).Conclusion:This study developed and validated a malignant probability prediction model of GGN based on 18F-FDG PET/CT imaging and clinical features.The model has good diagnostic efficiency and high specificity,which is helpful to improve the level of preoperative diagnosis of GGN.Part Ⅲ The diagnostic value of 18F-FDG PET combined with CT texture analysis in distinguishing malignant from benign ground-glass nodulesObjective:CT texture features are potentially radiological biomarkers in the differential diagnosis of lung cancer,prediction of tumor growth,gene expression and efficacy evaluation.In addition,PET/CT imaging has been recognized for its application in the field of lung cancer as a non-invasive dual-modal imaging that reflects tumor heterogeneity macroscopically.The purpose of this study was to establish a multimodal comprehensive prediction model based on CT texture parameters,semantic features,and PET metabolic parameters through analyzing the PET/CT images of patients with groundglass nodule(GGN)who underwent 18F-FDG PET/CT examination before operation and to investigate,and evaluate its diagnostic performance.Methods:Data of 165 patients who underwent 18F-FDG PET/CT examination in our hospital for indeterminate GGN from January 2012 to March 2020 were retrospectively collected.Propensity score matching(PSM)was performed to select GGNs with similar baseline characteristics.LIFEx software was used to extract 49 CT radiomic features,and the least absolute shrinkage and selection operator(LASSO)algorithm was used to select parameters and establish the radiomics score(Rad-score).Logistic regression analysis was performed combined with semantic features to construct a CT radiomics model,which was combined with SUVmax to establish the PET+CT radiomics model.Receiver operating characteristic(ROC)was used to compare the diagnostic efficacy of different models.Results:After PSM at 1:4,190 GGNs were divided into benign group(n=23)and adenocarcinoma group(n=92)according to postoperative pathology.After texture analysis,the Rad-score with three CT texture features was constructed for each nodule.Compared with the Rad-score and CT radiomics model(AUC:0.704(95%CI:0.562-0.845)and 0.908(95%CI:0.842-0.975),respectively),PET+CT radiomics model had the best diagnostic efficiency(AUC:0.940,95%CI:0.889-0.990),and there was significant difference between each two of them(P=0.001-0.030).Conclusion:SUVmax can effectively improve CT radiomics model performance in the differential diagnosis of benign and malignant GGNs.PET+CT radiomics might become a noninvasive and reliable method for differentiating of GGNs.Part Ⅳ Application of dual-stream 3D convolutional neural network based on 18F-FDG PET/CT in distinguishing benign and invasive adenocarcinoma in ground-glass lung nodulesObjective:Machine learning methods have been introduced into medical image analysis and there has been a clear shift towards the use of deep learning methods,especially multilayer convolutional neural network(CNN).Currently,there are rare reports on the application of three-dimensional CNN(3D-CNN)to differentiate benign from malignant GGN by analyzing PET/CT images and two-way CNN(data obtained from PET and CT as input streams,respectively).Therefore,this study aims to train,validate,and test the 3DCNN based on 18F-FDG PET/CT images and evaluate CT,PET,and PET/CT 3D-CNN performance in distinguishing benign lesions and IAC.Methods:The GGN patients who underwent 18F-FDG PET/CT examination in our hospital were retrospectively analyzed,and benign lesions and IAC were screened out according to postoperative pathology and clinical follow-up results.According to the ratio of 7:3,the data were randomly divided into training data and testing data.Partial image feature extraction software was used to segment PET and CT images,and the training data after using the data augmentation were used for the training and validation(fivefold crossvalidation)of the three CNNs(PET,CT,and PET/CT networks).The classification performance of different CNNs on training data and test data is evaluated according to the average accuracy,sensitivity and specificity.Results:A total of 23 benign nodules and 92 IAC nodules from 106 patients(115 GGNs)were included in this study.In the training set,the performance of PET network(accuracy,sensitivity,and specificity of 0.92±0.02,0.97±0.03,and 0.76±0.15)was better than the CT network(accuracy,sensitivity,and specificity of 0.84±0.03,0.90±0.07,and 0.62±0.16)(especially accuracy was significant,P-value was 0.001);in the testing set,the performance of both networks declined.However,the accuracy and sensitivity of PET network were still higher than that of CT network(0.76 vs.0.67;0.85 vs.0.70).For dual-stream PET/CT network,its performance was almost the same as PET network in the training set(P-value was 0.372-1.000),while in the testing set,although its performance decreased,the accuracy and sensitivity(0.85 and 0.96)were still higher than both CT and PET networks.Moreover,the accuracy of the PET/CT network was higher than that of two nuclear medicine physicians(Physician 1 with 3 years of work experience and Physician 2 with more than 5 years of work experience),with accuracies of 0.85,0.70,and 0.73,respectively.Conclusion:The 3D-CNN based on 18F-FDG PET/CT can be used to distinguish benign lesions and IAC in GGNs,and the performance is better when both CT and PET images are used together. |
PDF Full Text Request