Abstract:
Objective: To investigate the feasibility of CT radiomics analysis for the identification of exon 9 and 11 mutations of c-kit gene in gastrointestinal stromal tumor (GIST). Materials and Methods: We retrospectively analyzed 49 GIST enhanced CT cases with c-kit mutation confirmed by molecular pathology, 9 had exon 9 mutation, and 40 had exon 11 mutation. 105 radiomics features were extracted by using python pyradiomics, and feature selection was performed by LASSO regression. Then, the differences were compared and the identification model was constructed. Results: By LASSO selection, 7 valuable radiomics features were selected, and the logistic regression model had a good performance (area under curve, AUC=0.79). Conclusion: CT radiomics analysis had the potential to distinguish c-kit gene exon 9 and 11 mutations in GIST.