P13.16.A MACHINE LEARNING-BASED RADIOMIC MODEL USING MULTIPARAMETRIC MAGNETIC RESONANCE IMAGING FOR PREDICTION OF POSTOPERATIVE VISUAL RECOVERY OF PITUITARY ADENOMA PATIENTS

Abstract

Abstract BACKGROUND Preoperative prediction of the visual outcome after pituitary adenoma surgery is crucial to clinical decision-making but remains challenging. We aimed to develop radiomic models based on multiparametric magnetic resonance imaging (MRI) to predict the postoperative visual outcome for pituitary adenoma patients using machine learning approaches. MATERIAL AND METHODS A cohort of 152 patients who underwent endoscopic endonasal transsphenoidal surgery of pituitary adenomas were retrospectively enrolled and divided into the recovery group and the non-recovery group based on the six-month postoperative visual examinations. Radiomic features of the optic chiasm were extracted from preoperative T1-weighted imaging (T1WI), T2-weighted imaging (T2WI) and contrast-enhanced T1-weighted imaging (T1CE), respectively. Predictive models were constructed by least absolute shrinkage and selection operator wrapped with support vector machine through the five-fold cross-validation in the development cohort, and then evaluated in the independent test cohort. Area under the receiver operating characteristic curve (AUC), accuracy, specificity and sensitivity were used to assess the predictive performance of models. RESULTS A total of four predictive models were established based on selected radiomic features from three sequences individually or in combinations. Three features from T1WI, two features from T2WI and nine features from T1CE were selected in the multiparametric radiomic model that represented the best performances among the four models, with AUC values of 0.851 (95% CI: 0.768 - 0.933) in the development cohort and 0.847 (95% CI: 0.722 - 0.971) in the independent test cohort. CONCLUSION Our proposed machine learning-based radiomic model using multiparametric MRI can be utilized to assist in predicting the postoperative visual recovery of pituitary adenoma patients in clinical practice.