Prediction Of Diabetes Insipidus In Patients With Pituitary Macroadenoma After Surgery Based On MRI Imaging And Clinical Features

Objective:To explore the clinical and radiomics features of patients with pituitary macroadenoma,and to construct a model to predict the occurrence of postoperative diabetes insipidus.Methods:A total of 54 patients admitted to the Hospital from January 2017 to October2022 were retrospectively analyzed.They were diagnosed with pituitary adenoma by pathology,and diabetes insipidus was determined by calculation of urine volume or pituitrin use after surgery.Onekey software based on python,Free Statistics software based on R(V1.7.1)and SPSS 27.0 software were used to analyze the clinical data and brain Magnetic Resonance Imaging(MRI)images of patients.According to the inclusion and exclusion criteria,the data of 54 patients were collected and randomly divided into training set and validation set.To explore the risk factors of postoperative diabetes insipidus in patients with pituitary macroadenoma,namely clinical characteristics.Then ITK-SNAP software was used to delineate the Region of Interest(ROI)in the image,and Py Radiomics was used to extract the radiomics features from the image.Mann-Whitney U test statistical test and feature selection were performed on all radiomics features.Appropriate radiomics features were selected,and then combined with clinical features,a prediction model was constructed to predict the occurrence of postoperative diabetes insipidus in patients with pituitary macroadenoma.Hosmer-Lemeshow test was used to evaluate the agreement between the predicted results and the actual results.Decision curve analysis was performed to evaluate the effectiveness of the radiomics model and the combination model of clinical features and radiomics features.Result:Of the 54 patients included,31 were male and 23 were female.The male to female ratio was about 1.34:1.There were 14 patients who did not develop diabetes insipidus and 40 patients who developed diabetes insipidus after surgery.T use independent sample test,the Mann-Whitney inspection or 2 to compare clinical features,the results showed:leaf spot and after T1 sequence of pituitary tumor growth direction associated with postoperative diabetes insipidus,with gender,age,tumor diameter and degree of tumor resection.Diabetes insipidus is more likely to occur when there is no posterior pituitary bright spot in the T1 sequence and then when the tumor breaks through the sellar septum and grows.Then,combined with the selected clinical features(the posterior pituitary highlight on T1 sequence and the penetration of the sellar septum of the tumor),the clinical feature-radiomics feature combination model was constructed.In the training set,the AUC value of the model combined with radiomics features alone was 0.909(95%CI,0.777-1.000),and the AUC value of the clinical-radiomics combined model was 0.885(95%CI,0.727-1.000).In the validation set,the AUC value was 0.892(95%CI,0.723-1.000)for the model combined with radiomics features only,0.923(95%CI,0.786-1.000)for the model combined with clinical features and radiomics features,and0.923(95%CI,0.786-1.000)for the model combined with clinical features and radiomics features.In conclusion,the radiomics feature model and the clinical feature-radiomics feature combination model have similar effects in predicting diabetes insipidus,and have predictive value for the occurrence of diabetes insipidus in patients with pituitary macroadenoma after surgery.Conclusion:The occurrence of diabetes insipidus in patients with pituitary adenoma after surgery is related to the bright spots in the posterior pituitary lobe.Combining clinical features with imaging features,it can be used to predict the occurrence of diabetes insipidus in patients with pituitary macroadenoma before surgery,providing evidence for the postoperative management of patients with pituitary macroadenoma.