Prediction of the consistency of large pituitary adenoma based on CT density combined with texture parameter modeling / 中国医学影像技术

ABSTRACT

Objective: To explore the value of CT density combined with texture parameters based on CT plain image in predicting the consistency of large pituitary adenoma. Methods: Totally 50 patients with large pituitary adenoma confirmed by operation and pathology were enrolled and divided into soft group (n=30) and hard group (n=20) according to intraoperative pituitary consistency. The largest slice of the tumor on the CT image was selected, then ROI was manually outlined, CT value of the lesion was measured, and the texture feature parameters were extracted. CT values and texture features were compared between the two groups. Multivariate Logistic regression analysis was used to analyze the variables, and the model for predicting the pituitary adenoma consistency was established. ROC curve was drawn to evaluate its predictive value. Results: There was statistically significant difference in CT value between soft group and the hard group (P=0.031), and AUC in predicting tumor consistency was 0.662. A total of 77 texture parameters were extracted based on plain CT images, and 4 texture parameters were found with statistically significant differences between the two groups, including the Quantile 90, inertia, variance and contrast, with AUC of 0.662, 0.663, 0.672 and 0.663, respectively. AUC of texture feature model established with multivariate Logistic regression analysis in predicting the pituitary adenoma consistency was 0.690, of CT value combined with the texture parameter model was 0.782. Conclusion: The model established with CT value combined with texture parameters has high value in predicting the pituitary adenoma consistency, which is helpful to clinical selection of surgical plans.