The Predictive Value Of Preoperative Conventional MRI Findings And Texture Analysis For Postoperative Prognosis Of Functional Pituitary Adenoma

Part I Investigate the Relationship between Preoperative Conventional MRI Findings and Postoperative Prognosis of Functional Pituitary AdenomaObjective To explore the value of preoperative MRI findings in predicting total tumor resection and biochemical response in patients with functional pituitary adenoma.Methods Preoperative and postoperative MRI and clinical data of 75 patients with functional pituitary adenoma in our hospital from December 2016 to June 2021 were retrospectively collected.The signal intensity of different sequences,tumor maximum size,cystic degeneration and hemorrhage,Konsp grading obtained from preoperative MRI were recorded.Then,Analyze whether there is a statistical difference in the imaging findings between the total tumor resection group and the non-total tumor resection group,the biochemical remission group and the non-remission group after surgery.Statistical methods including Chi-square test,Fisher’s exact probability test and T test.Results In 75 cases of functional pituitary adenoma,52 cases（69.33%）achieved total tumor resection and 51 cases（68.00%）achieved biochemical remission.The diameter of the completely resected pituitary adenoma was smaller than that of the incomplete resected pituitary adenoma（20.17±9.29 mm vs 25.96±10.33mm）,with statistically significant differences（Z=2.402,P < 0.05）.Pituitary adenomas with marked and moderate enhancement on MRI were less likely to achieve total tumor resection and biochemical remission after surgery than those with mild enhancement,with statistical significance differences（P < 0.05）.The pituitary tumors with Knosp grades 0-2 were more likely to achieve total tumor resection and biochemical remission after surgery than those pituitary tumors with Knosp grades 3-4,with statistically significant differences（P < 0.01）.However,There were no significant differences in the distribution of T1 WI and T2 WI signal intensity between the total resection group and the non-total resection group,the biochemical remission group and the non-remission group,as well as the distribution of intratumor cystic degeneration and hemorrhage.（P > 0.05）.Conclusion Preoperative tumor size,Knosp grade and MRI enhanced sequence of patients with functional pituitary adenoma has certain predictive value for postoperative tumor resection degree and biochemical response degree.Part II Investigate the Predictive Value of Preoperative MRI Texture Analysis for Postoperative Resection Degree of Functional Pituitary AdenmaObjective To explore the predictive value of preoperative MRI texture analysis on whether functional pituitary adenoma can be completely resected.Methods Patients with good image quality and a diameter greater than 1cm in the Part I were selected as the research objects.A total of 57 functional pituitary adenoma patients were selected.Among them,43 patients achieved total tumor resection and 14 patients failed to achieve total tumor resection.Tumor boundaries in coronal T2 WI and coronal CE-T1 WI were delineated using the third-party software.Then,the AK.software was used to extract texture features from the region of interest,including first-order features,gray level co-occurrence matrix features,shape features and gray level run length matrix features.The difference of all texture parameters between the complete resection group and the incomplete resection group was compared.The ROC curve was drawn for the texture features with statistically significant differences to evaluate their discriminating performance,and then the Logistic forward stepwise regression method was used to further screen the above-mentioned texture features with statistically significant differences and a logistic regression prediction model was constructed.The ROC curve of the model was drawn to evaluate the discriminative power of the model.Results Twenty-one texture features extracted from T2 WI sequence have diagnostic value,and Run Length Non Uniformity of the gray level run length matrix features has the highest diagnostic efficiency,with an area under the curve of 0.761,sensitivity of85.7% and specificity of 62.8%.Nineteen texture features extracted from CE-T1 WI sequence have diagnostic value,and Major Axislength of the Shape feature has the highest diagnostic efficiency,with an area under the curve of 0.777,sensitivity of 85.7%and specificity of 65.1%.Four texture features were screened out by Logistic forward stepwise regression,including T2WI-GLCM-Correlation 、CE-T1WI-Shape-Major Axis Length 、 CE-T1WI-Shape-Maximun2 DDiameter Slice 、CE-T1WI-Shape-Maximum3 DDiameter,after incorporated into the Logistic regression model and drew the ROC curve.The discriminative power of the model was an area under the curve of 0.899,sensitivity of 81.4%,specificity of 100% and accuracy of80.7%.Conclusion Preoperative MRI texture analysis has certain predictive value for postoperative resection degree of functional pituitary tumor.