| Lymph node metastasis is one of the main metastatic routs of cervical cancer and is the main risk factor for patients’ prognosis.Lymph node enlargement(short diameter>1 cm)is the main criterion for imaging diagnosis of lymph node metastasis,but cancer involvement is often found in normal-sized lymph nodes(short diameter ≤1 cm)in clinical practice.It has been reported that more than 50%of the metastatic lymph nodes are normal-sized lymph nodes.However,there were few studies showed the proportion of cervical cancer patients with normal-sized metastatic lymph nodes.Most patients with normal-sized metastatic lymph nodes are difficult to be accurately diagnosed by traditional imaging methods before operation,and therefore receive surgery-based comprehensive treatment,but the prognosis of such patients has not been evaluated separately.Therefore,the first purpose of this study is to investigate the proportion and oncology prognosis of patients with normal size lymph node metastasis,so as to get more attention in normal-sized metastatic lymph nodes of cervical cancer.Normal-sized lymph node metastasis is not uncommon in clinic,but the ability of preoperative diagnosis of normal-sized lymph node metastasis is low,mainly because the conventional imaging mainly depends on the shape of lymph nodes to evaluate whether they have metastasis,and as a result,normal-sized metastatic lymph nodes are often missed.Radiomic can use the high-dimensional image information which is difficult to be recognized visually to evaluate the normal-size lymph nodes in order to accurately diagnose the normal-size lymph node metastasis before operation.Therefore,the second purpose of our study is to use radiomic methods to extract CT imaging features of pelvic or abdominal lymph nodes,and to construct and validate a point-to-point radiomic model for predicting normal-sized lymph node metastasis in cervical cancer,so as to improve the diagnostic efficiency of normal-sized metastatic lymph nodes,and help to determine patients staging,and in addition,to guide the formulation of individualized treatment strategy.Chapter 1 Investigation and Prognostic Analysis of Normal-sized Lymph Node Metastases in Cervical CancerSection 1 Investigation of normal-sized lymph node metastases in cervical cancer[Objective]To analyze the proportion of patients with normal size lymph node metastasis in early cervical cancer.[Methods]The CT images of 1921 cervical cancer patients were retrospectively analyzed,and the short diameter of pelvic and abdominal lymph nodes was measured in CT cross-sectional images.The short-axis diameter of the lymph node≤1cm were defined as normal.Taking the postoperative pathology of the lymph nodes as the gold standard,the proportion of patients with normal-sized lymph nodes metastasis in the total nodal metastasis patients was analyzed.[Results]A total of 1553 patients with stage IA1(LVSI+)-IIA2(2009FIGO stage)were included in this section,among these,340 patients with pathologically confirmed nodal metastasis.Combined with evaluation of CT imaging,a total of 235 patients were normal-sized lymph node metastases.The proportion of normal-sized lymph node metastasis patients was 69.1%.[Conclusion]69.1%of early stage cervical cancer patients with nodal metastasis were normal-sized lymph node metastasis.Attention should be paid to normal-sized metastatic lymph nodes.Section 2 Prognostic analysis of cervical cancer patients with normal-sized lymph node metastasis[Objective]To analysis of prognosis in cervical cancer patients with normal-sized lymph node metastasis.[Methods]Kaplan-Meier method was used to compare the 5-year OS and DFS of the three groups.The Cox proportional hazard regression model was used to determine the independent prognostic factors.The subgroup analysis further compared the OS and DFS between patients with enlarged lymph node metastasis and normal-sized lymph node metastasis after PSM matching.[Results]A total of 1264 patients were included,of which 1071 were in the non-metastatic group,134 were in the normal-sized metastatic group,and 59 were in the enlarged metastatic group.The 5-year OS of the three groups were 91.1%,73.7%,and 45.4%,respectively.The 5-year DFS was 90.3%,75.5%,and 50.5%,respectively.P<0.050 for each pairwise comparison in OS and DFS.In the subgroup analysis,the normal-sized metastatic group included 71 patients and the enlarged metastatic group included 44 patients after matching.The normal-sized metastatic group had better OS and DFS at 5 years than the enlarged and metastatic group(OS:78.9%vs.41.2%,P=0.017;DFS:74.2%vs.53.1%,P=0.071).The risk of death and recurrence in the enlarged metastatic group were 4.462 and 3.118 times than that in the normal-sized metastatic group,respectively(P<0.050).[Conclusion]For patients with early staged cervical cancer,the oncological outcomes of patients with normal-sized lymph node metastasis were worse than those without lymph node metastasis,so attention should be paid to the diagnosis of normal-sized metastatic lymph nodes before operation.Enlarged lymph node metastasis has a higher risk of death and recurrence than normal-sized lymph node metastasis.For patients with enlarged metastatic lymph nodes,surgery-based comprehensive treatment should be avoided.Chapter 2 CT based radiomic model for point-to-point prediction of normal-sized lymph node metastasis in cervical cancer[Objective]We aimed to develop and validate a CT based radiomic model for point-to-point prediction of normal-sized lymph node metastasis(LNM)in cervical cancer.[Methods]A total of 273 LNs of 219 patients from 12 centers were enrolled in this study.We randomly divided the LNs from 2 centers into training and internal validation cohorts,and the rest as external validation cohort.Radiomic features were extracted from arterial and venous phase CT images,respectively.We used two different approaches to build ten single-phase models and five combined models.One used the mRMR for feature selection and 4 different classifiers(logistic,SVM,DT,RF)to construct the models.Another utilized all the features to train an artificial neural network(ANN)and built models.The performance of all models was accessed with the area under receiver operating characteristic(ROC)curves(AUC).[Results]There were 15 radiomic models we built.Among these models,the performance of the combined models surpassed that of the single-phase models.The combined ANN model achieved the highest AUCs of 0.912 and 0.859 in training and internal validation cohorts,respectively,while in the external validation cohort,the AUC value was 0.800.[Conclusion:]We constructed a radiomic model incorporating the arterial and venous CT features for noninvasively point-to-point prediction of normal-sized LNM in cervical cancer patients.The application of the radiomic model could optimize clinical staging and decision-making. |
PDF Full Text Request