基于CT特征构建鉴别胃神经鞘瘤与胃肠道间质瘤的诊断模型  

A CT Features-based Prediction Model for Differentiating Gastric Schwannomas from Gastrointestinal Stromal Tumor

作　　者：凌文峰[1] 黄瑞滨[2] 戴卓智 张志强[1] 杨志企[1,5] LING Wen-feng;HUANG Rui-bing;Dai Zhuo-zhi;ZHANG Zhi-qiang;YANG Zhi-qi(Meizhou People's Hospital,Meizhou 514031,Guangdong Province,China;The First Affiliated Hospital of Shantou University Medical College,Shantou 515000,Guangdong Province,China;Shantou Central Hospital,Shantou 515031,Guangdong Province,China;Department of Radiology,Sun Yat-sen Memorial Hospital,Sun Yat-sen University,Guangzhou 528406,Guangdong Province,China;Guangdong Engineering Technological Research Center of Clinical Molecular Diagnosis and Antibody Drugs,Meizhou 514031,Guangdong Province,China)

机构地区：[1]梅州市人民医院,广东梅州514031 [2]汕头大学医学院第一附属医院,广东汕头515000 [3]汕头市中心医院,广东汕头515031 [4]中山大学附属孙逸仙纪念医院放射科,广东广州528406 [5]广东省临床分子诊断与抗体药物工程技术研究中心,广东梅州514031

出　　处：《中国CT和MRI杂志》2023年第12期146-148,共3页Chinese Journal of CT and MRI

基　　金：梅州市社会发展科技计划项目(2022B15);广东省医学科研基金项目(B2023445);国家自然科学基金项目(82101985)。

摘　　要：目的探讨基于CT定量和定性特征的模型鉴别胃神经鞘瘤和胃肠道间质瘤。方法收集116例胃肠道间质瘤和34例神经鞘瘤患者CT资料,收集CT定性特征包括形态、位置、生长方式、密度、边界、黏膜面溃疡、瘤内出血、坏死、囊变、钙化和定量特征包括长径、短径、病灶各期CT值并计算肿瘤各期强化程度、相对/绝对强化率。采用t检验或χ^(2)检验比较胃肠道间质瘤和胃神经鞘瘤间CT特征差异。采用多因素逻辑回归筛选独立危险因子。采用ROC曲线评估模型的预测效能。结果对比胃肠道间质瘤,胃神经鞘瘤多发生在胃体,呈腔外型、混合型生长,黏膜面溃疡多见,且瘤体积较大,肿瘤静脉期CT值、静脉期强化程度、静脉期绝对强化率较高。多因素逻辑回归分析显示肿瘤长径、位置、生长方式、边界、囊变和静脉期绝对强化率是鉴别胃神经鞘瘤与胃肠道间质瘤的独立危险因子。基于联合CT定量和定性特征的联合模型曲线下面积均高于基于CT定量特征模型(AUC 0.955 vs.0.837,P=0.003)和基于CT定性特征模型(AUC 0.955 vs.0.896,P=0.009)。结论肿瘤长径、位置、生长方式、边界、囊变和静脉期绝对强化率有助于神经鞘瘤与胃肠道间质瘤鉴别诊断,基于上述CT特征建立的联合模型能够较好鉴别两者。Objective To explore the value of CT quantitative and qualitative features-based models in distinguishing between gastric schwannoma(GS)and gastrointestinal stromal tumor(GIST).Methods This retrospective study analyzed the CT findings of 116 patients with GIST and 34 patients with GS.The qualitative CT features(e.g.,morphology,location,growth pattern,density,boundary,mucosal surface ulcer,intratumoral bleeding,necrosis,calcification,cystic degeneration)and quantitative CT features(e.g.,maximum diameter,minimum diameter,and CT value of lesion)were collected.The enhancement degree and relative/absolute enhancement rate of lesion in each phase were calculated.The t-test or chi-squared test was used to compare the CT features between GIST and GS groups.Multivariate logistic regression was used to identify independent risk factors and build a prediction model,which was evaluated by the ROC curve.Results Patients with GS were more likely to have tumors in the stomach body,extraneous and mixed growth patterns,a higher incidence rate of mucosal surface ulcers,and larger tumor sizes than those with GIST.The CT value,enhancement degree,and absolute enhancement rate of the venous phase were higher in GS patients than that in GIST patients.Multivariate logistic regression analysis showed that tumor length,location,growth pattern,boundary,cystic degeneration,and absolute enhancement rate at the venous phase were independent risk factors for distinguishing GS from GIST.The model that combined both CT quantitative and qualitative features demonstrated superior performance in differentiating between GS and GIST,with a higher area under the curve compared to the models based solely on CT quantitative features(AUC 0.955 vs.0.837,P=0.003)or CT qualitative features(AUC 0.955 vs.0.896,P=0.009).Conclusion The tumor maximum diameter,location,growth pattern,boundary,cystic degeneration,and absolute enhancement rate of the venous phase are helpful in the differential diagnosis of GS and GIST.The combined model based on these CT features

关 键 词：胃神经鞘瘤 胃肠道间质瘤 CT特征 

分 类 号：R737.2[医药卫生—肿瘤] R814.42[医药卫生—临床医学]