Objective  To explore the clinical application value of a machine learning (ML) model based on enhanced computed tomography (CT) radiomics in differentiating low-risk gastric stromal tumors (GST) from benign gastric submucosal tumors (GSMT).

  Methods  The clinical and imaging data for 84 cases of low-risk GST and 51 cases of benign GSMT from Yijishan Hospital Affiliated to Wannan Medical College were retrospectively analyzed. Patients were randomly assigned into a training set (n=94) and a test set (n=41). Based on enhanced CT vein phase, ITK-SNAP software was used to segment the images; AK software was used to extract the image group characteristics; mRMR, Spearman rank correlation, and LASSO regression were used to reduce the dimensions of the features; and the image group label score (Radscore) was established. Single- and multi-factor Logistic regression was used to screen independent risk factors. Support vector machine (SVM) was used to establish a prediction model, and the test set was used to test the generalization ability of the model. The area under the receiver operating characteristic (ROC) curve (AUC) of the subjects was used to evaluate the effectiveness of the model.

  Results  Multivariate analysis showed that age, shape, location, long diameter/short diameter (LD/SD), and Radscore were independent risk factors. The AUC for training set and test set of the prediction model established by SVM were 0.933 and 0.913, respectively.

  Conclusions  The model, based on SVM algorithm of enhanced CT radiomics, was able to effectively identify low-risk GST and benign GSMT with good generalization ability.