Fig. 1

The process of predictor selection. In the initial XGBoost model, we used all clinical data before surgery to construct the model. a Shows the importance of each feature. It’s easy to find out that most hematological test features don’t have an important impact on GIST diagnosis except ALT. Similarly, we built the second model using the data only from enhanced CT, endoscopy and EUS (b). The latter three features also have little importance, and that’s why they were excluded. The first 6 predictors were determined to be the most important features and utilized in the final model development. The importance of the six predictors in the final XGBoost model are shown in c. The most important predictors of this model is the existence of liquid area inside the tumor under EUS, following by the ratio of long and short diameter under CT, the CT value of the tumor, the enhancement of the tumor in arterial period and venous period, existence calcified area inside the tumor under EUS. (The specific values and 95 CI of importance are shown in Additional file 2: Table S1.)