| PART 1 Potential value of FDG PET/CT in predicting occult lymph node metastasisin clinical stage IA lung adenocarcinomaObjective To evaluate whether PET/CT parameters of primary tumor can predict occult lymph node metastasis in clinical stage IA lung adenocarcinoma.Methods From October 2006 to September 2015,272 patients with clinical stage IA lung adenocarcinoma were retrospectively reviewed.All patients underwent preoperative 18F-FDG PET/CT scan followed by lobectomy and systematic lymph node dissection.The correlation between occult lymph node metastasis and PET/CT parameters(SUVmax,MTV,TLG)of primary tumor as well as other clinicopathological factors were analyzed.Results Occult lymph node metastasis was detected in 50 patients(18.4%),24(8.8%)with N1 involvement and 26(9.6%)with N2 involvement.Among the 272 patients enrolled(39 with pure ground glass nodule,59 with part-solid nodule,174 with solid nodule),all patients with pure ground glass nodule or nodule?1cm were pNO.For the 233 part-solid and solid nodule,primary tumor SUVmax,MTV1.5,TLG1.5,nodule type,tumor location,histological grade,visceral pleural invasion were associated with lymph node metastasis(P<0.05).Multivariate analysis revealed that SUVmax>2.41,MTVi.5>1.05cm3,TLG1.5>2.04,central location and higher histological grade were independent predictors of occult lymph node metastasis.Visceral pleural invasion was significantly associated with occu.lt lymph node metastasis for model including MTV1.5 or TLG1.5(P<0.05),but not statistically significant for model including SUVmax(P=0.055).By using ROC analysis,the area under the curve(AUC)for model including SUVmax was slightly higher than that including MTV1.5 or TLG1.5.With a cut-off value of 2.41 in primary tumor SUVmax,the sensitivity,specificity and negative predictive value were 90.0%,61.7%and 95.8%,respectively,and the AUC was 0.761(95%CI,0.700?0.823).Conclusions For clinical stage IA lung adenocarcinoma,PET/CT parameters of primary tumor,central location and histological grade were independent risk factors for occult lymph node metastasis.Patients with visceral pleural invasion had a higher rate to develop occult lymph node metastasis.Less invasive treatment were possible alternatives in clinical NO lung adenocarcinoma with pure ground glass density,nodule?1cm or SUVmax?2.41 due to the very low risk of nodal involvement.PART 2 PET/CT-based radiomics signature for predicting occult lymph nodemetastasis in clinical stage I lung adenocarcinomaObjective To evaluate the usefulness of radiomics signature derived from PET/CT in predicting occult lymph node metastasis in clinical stage I lung adenocarcinoma.Methods We retrospectively reviewed clinical stage I lung adenocarcinoma underwent surgery at CICAMS from October 2006 to December 2015.All patients underwent preoperative 18F-FDG PET/CT and breath hold chest CT scan followed by lobectomy and systematic lymph node dissection.A matched case-control study was performed among 130 patients.Patients with occult lymph node metastasis were taken as cases.The controls were matched according to a 1:1 ratio.Variables used for matching were T stage and histological grade.Radiomics features of primary tumor were extracted from three examined modalities(PET,free-breathing CT and breath-hold chest thin-section CT).Three prediction models were built using radiomics signatures of PET(Model 1),PET combined with free-breathing CT(Model 2)and PET combined with breath-hold chest thin-section CT(Model 3),respectively.The predictive ability of the models and clinical candidate predictors was assessed both in the training and validation set.Results Multivariable analysis on the correlation between occult lymph node metastasis and SUVmax of primary tumor as well as other clinical factors demonstrated that SUVmax was the only independent risk factor of occult lymph node metastasis.With a cut-off value of 2.41 in primary tumor SUVmax,the sensitivity and specificity for predicting occult lymph node metastasis were 90.8%and 58.5%,respectively,and the area under the curve(AUC)was 0.806(95%CI,0.731?0.880).As for the radiomics models,the AUC for three models in the training cohort were 0.890(95%CI,0.825?0.955),0.928(95%CI,0.878?0.978)and 0.953(95%CI,0.912?0.993),respectively.The AUC in the validation cohort were 0.787(95%CI,0.641?0.934,sensitivity 80.0%,specificity 75.0%),0.855(95%CI,0.734?0.976,sensitivity 100.0%,specificity 65.0%)and 0.917(95%CI,0.820?1.000,sensitivity 90.0%,specificity 90.0%),respectively.Conclusion The AUC of Model 2 and Model 3 were higher than SUVmax in predicting occult lymph node metastasis in the validation cohort.The model derived from radiomics signatures of PET combined with breath-hold thin-section CT demonstrated the best prediction performance. |
PDF Full Text Request