A Multimodal Model Fusing Multiphase Contrast-Enhanced CT And Clinical Characteristics For Predicting Lymph Node Metastases Of Pancreatic Cancer

Background:Pancreatic cancer is the most malignant cancer of its kind;its mortality is almost that of its morbidity.Surgical resection remains the only curable treatment.However,many high-risk factors exist preoperatively that hinder the operative prognosis,among which lymph node metastasis(LNM)is prominent.Contrast-enhanced Computed Tomography(CECT)can assist clinicians in evaluating the lymph node(LN)status of pancreatic cancer preoperatively.However,the accuracy of diagnostic results is often not satisfied.Therefore,it is of clinical significance to improve the accuracy of preoperative diagnosis of LNM in pancreatic cancer.Objective:To develop a multimodal model that combines multiphase of CECT imaging and clinical characteristics to preoperatively predict lymph node metastasis in pancreatic cancer patients.Methods:We first proposed a new classifier fusion strategy(CFS)based on new evidential reasoning(ER)rule(CFS-nER)by combining nomogram weights into a previous ER rule based CFS for better performance.And three kernelled support tensor machine(KSTM)-based classifiers with plain,arterial and venous phases of CECT as the inputs,respectively,were constructed.They were then fused based on CFS-nER to construct fusion model of multiphase CECT for integrating the information from each phase of CECT.The clinical characteristics were analyzed by univariate logistic regression(LR)and multivariable LR to screen risk factors,which were used to construct correspondent risk factor-based classifiers,respectively.Finally,the fusion model of the three phases of CECT and each risk factor-based classifier were fused further to construct the MMM based on our CFS-nER,named MMM-nER.This retrospective study consisted of 186 patients diagnosed with pancreatic cancer between June 2010 and July 2020 from four clinical centers in China,88(47.31%)of whom had LNM.The data from two centers was used to develop predictive models with five-fold cross validation technique and the data from another two centers was used to perform independent test.Results:The fusion model of the three phases of CECT imaging performed better overall than single and two-phase fusion models;this implies that the three considered phases of CECT were supplementary and complemented one another.Our MMM-nER further improved the predictive performance,which implies that our CFS-nER can complement availably the supplementary information between CECT and clinical characteristics.The MMM-nER had better predictive performance than based on previous CFSs,which presents the advantage of our CFS-nER.Conclusion:We proposed a new CFS-nER,based on which the fusion model of the three phases of CECT and MMM-nER were constructed and performed better than all compared methods.MMM-nER based on our CFS-nER achieved an encouraging predictive performance,implying that it can assist clinicians in noninvasively and preoperatively evaluating the lymph node status of pancreatic cancer.