Computed-Tomography-Based Radiomics Model for Predicting the Malignant Potential of Gastrointestinal Stromal Tumors Preoperatively: A Multi-Classifier and Multicenter Study

Background: Our goal was to establish and verify a radiomics risk grading model for gastrointestinal stromal tumors (GISTs) and to identify the optimal algorithm for risk stratification. Methods: We conducted a retrospective analysis of 324 patients with GISTs, the presence of which was confirmed by surgical pathology. Patients were treated at 3 different hospitals. A training cohort of 180 patients was collected from the largest center, while an external validation cohort of 144 patients was collected from the other two centers. To extract radionomic features, regions of interest (ROIs) were outlined layer by layer along the edge of the tumor contour on CT images of the arterial and portal venous phases. The dimensionality of radiomic features was reduced, and the top 10 features were ranked in order of importance before modeling. The training cohort used three classifiers (logistics regression, support vector machine(SVM),and random forest) to establish three GIST risk classification prediction models, and a receiver operating characteristic curve (ROC) and was used to compare model performance, which was validated using external data. Results: In the training and external validation cohorts, the area under the curve (AUC) of the logistic regression was 0.84 and 0.85, respectively. SVM was 0.81 for the training and 0.80 for the external validation cohort, while random forest was 0.88 for the training and 0.90 for the external validation cohort. The random forest model performed the best in both the training and the external validation cohort and was generalizable. Conclusion: Different machine learning models based on CT radiomics can predict the overall risk of GISTs. Among these, the random forest algorithm has the highest prediction efficiency and could be readily generalizable, which we confirmed using external validation data. This model has the potential to help clinicians predict the overall GIST risk level prior to surgery.