Application Value Of CT Texture Analysis In Distinguishing Gastrointestinal Stromal Tumor From Other Mesenchymal Tumors

Application of CT texture analysis in differentiating gastric stromal tumors from other mesenchymal tumorsPART 1Objective: To test the feasibility of differentiate gastric stromal tumors from other mesenchymal tumors using texture analysis based on contrast-enhanced CT images.Methods: The contrast-enhanced CT image data of 74 patients with gastric stromal tumor and 30 patients with other mesenchymal tumor confirmed by postoperative pathology were retrospectively analyzed.First,two senior radiologists read CT images to obtain a subjective CT signs model,including mean diameter,presence or absence of peripheral lymph nodes,growth pattern,delayed enhancement,stomach tumor plain scan and enhanced CT value.Secondly,two radiologists used ITK-SNAP software to manually segment gastric mesenchymal tumors in CT images,extract radiology features,and construct radiology features models.Finally,a diagnostic model integrated with subjective CT signs and radiomics signatures was constructed.Finally,a diagnostic model combining subjective CT signs and radiological characteristics was established.The diagnostic efficacy of three models in differentiating gastric stromal tumors from other mesenchymal tumors was compared by using receiver operating characteristic curves(ROC).Results: There are statistically significant differences between the gastric stromaltumors and other mesenchymal tumors in the mean diameter,delayed enhancement,growth pattern,peripheral lymph nodes,plain scan density and radiomics signature(p <0.05).While the combined model yields the highest area under ROC curve(AUC)value(0.918),sensitivity(96.00%)and specificity(89.2%)among the three models.Conclusion: A diagnostic model combining subjective CT signs and radiological features can improve the diagnostic accuracy of gastric mesenchymal tumors.PART 2 Predictive value of preoperative risk classification of gastric stromal tumor based on CT image texture analysisObjective:To explore the predictive value of CT texture analysis for preoperative risk classification of gastric stromal tumors.Methods: A total of 117 cases of gastric stromal tumors confirmed by surgery and pathology were selected and divided into extremely low / low risk group(n = 45)and medium / high risk group(n = 72)according to postoperative pathological results.First,CT images were read by two senior radiologists to acquire subjective CT signs model,including tumor site,mean diameter,plain and enhanced three-phase CT values.Then,the manual segmentation of gastric stromal tumors from the CT images was performed by the two radiologists to extract radiomics features via ITK-SNAP software,and to construct radiomics signature model.Finally,a diagnostic model integrated with subjective CT signs and radiomics signatures was constructed.The diagnostic efficacy of three models in differentiating extremely low / low risk group from medium / high risk group was compared by using receiver operating characteristic curves(ROC).Results: There are statistically significant differences between the extremely low / low risk group and medium / high risk group in the plain scan density,mean diameter and radiomics signature(p < 0.05).The area under ROC curve(AUC),sensitivity of subjective CT signs model was the lowest among the three models.While the combined model yields the highest AUC value(0.935),sensitivity(91.7%)and specificity(82.2%)among the three models.Conclusion: A diagnostic model combining subjective CT signs and radiological features can help accurately predict preoperative risk grading of gastric stromal tumors.