Preoperative prediction of Lauren classification in gastric cancer: a radiomics model based on dual-energy CT iodine map

Insights into Imaging

Original Article

Chongfei Ma¹  Xiaosheng Xu¹  Li Yang¹  Junyi Sun¹  Yang You¹  Min Li¹  Hongtao Qin²  Xianbo Yu³ 

[1] Department of Computed Tomography and Magnetic Resonance, Fourth Hospital of Hebei Medical University, No. 12, JianKang Road, 050010, Shijiazhuang, Hebei Province, People’s Republic of China;Department of Radiology and Nuclear Medicine, The First Hospital of Hebei Medical University, No. 89, Donggang Road, 050031, Shijiazhuang, Hebei Province, People’s Republic of China;Siemens Healthineers Ltd., 7, Wangjing Zhonghuan Nanlu, 100102, Beijing, People’s Republic of China;

关键词: Gastric cancer;    Lauren classification;    Dual-energy CT;    Iodine map;    Radiomics;   

DOI  :  10.1186/s13244-023-01477-8

 received in 2023-02-20, accepted in 2023-07-03,  发布年份 2023

来源: Springer

【 摘 要 】

ObjectiveTo investigate the value of a radiomics model based on dual-energy computed tomography (DECT) venous-phase iodine map (IM) and 120 kVp equivalent mixed images (MIX) in predicting the Lauren classification of gastric cancer.MethodsA retrospective analysis of 240 patients undergoing preoperative DECT and postoperative pathologically confirmed gastric cancer was done. Training sets (n = 168) and testing sets (n = 72) were randomly assigned with a ratio of 7:3. Patients are divided into intestinal and non-intestinal groups. Traditional features were analyzed by two radiologists, using logistic regression to determine independent predictors for building clinical models. Using the Radiomics software, radiomics features were extracted from the IM and MIX images. ICC and Boruta algorithm were used for dimensionality reduction, and a random forest algorithm was applied to construct the radiomics model. ROC and DCA were used to evaluate the model performance.ResultsGender and maximum tumor thickness were independent predictors of Lauren classification and were used to build a clinical model. Separately establish IM-radiomics (R-IM), mixed radiomics (R-MIX), and combined IM + MIX image radiomics (R-COMB) models. In the training set, each radiomics model performed better than the clinical model, and the R-COMB model showed the best prediction performance (AUC: 0.855). In the testing set also, the R-COMB model had better prediction performance than the clinical model (AUC: 0.802).ConclusionThe R-COMB radiomics model based on DECT-IM and 120 kVp equivalent MIX images can effectively be used for preoperative noninvasive prediction of the Lauren classification of gastric cancer.Critical relevance statementThe radiomics model based on dual-energy CT can be used for Lauren classification prediction of preoperative gastric cancer and help clinicians formulate individualized treatment plans and assess prognosis.Graphical abstract

【 授权许可】

CC BY   
© The Author(s) 2023

【 图 表 】

Fig. 8

Fig. 2

Fig. 18

Fig. 1

Fig. 2

Fig. 13

Fig. 1