Clinical and Radiomic-Derived Features Predict Hyponatremia After Pituitary Adenoma Resection

Background: Postoperative hyponatremia is a common complication of transsphenoidal pituitary adenoma resection. Symptoms range in severity, with seizures representing a serious complication. This study applies clinical and image-based radiomic analysis to predict the onset of postoperative hyponatremia. Methods: We conducted a single center, retrospective review of 77 patients from 2012 to 2024 who underwent transsphenoidal resection of pituitary adenomas. Preoperative and outcome variables were obtained. T1- and T2-weighted images were acquired as part of routine preoperative care. Scans were obtained by 1.5T or 3T scanners using standard pituitary protocols, and image slice thickness ranged from 1–3mm. Exact acquisition parameters varied based on scanner and year of acquisition, and outputs were studied for variations in quality (e.g., slice thickness). Masks were segmented or verified by a neuroradiology fellow (DA) near the completion of fellowship training. Radiomic feature extraction was performed using PyRadiomics (version 3.0.1), with classes that include first-order features, shape, gray level run length, gray level size zone matrix, gray tone differences, and gray level dependence matrix (GLDM). Least Absolute Shrinkage and Selection Operator (LASSO) regression was implemented. Results: 8 (10.4%) patients of the 77 developed postoperative or delayed hyponatremia after transsphenoidal resection. LASSO regression achieved a mean area under the curve (AUC) of 0.883 ± 0.079 and mean accuracy of 0.843 ± 0.036. 243 total features were used, with 26 nonzero features selected for prediction. The strongest positive predictors of hyponatremia included age at encounter (β = 1.54), uninsured/self-pay insurance status (β = 0.84), and wavelet-based GLDM dependence non-uniformity (β = 0.73), reflecting texture heterogeneity with higher values. Top negative predictors included first-order skewness (β = -1.35), maximum 2D diameter (β = -0.30), and first-order kurtosis (β = -0.57), which reflects flatter intensity distributions. Conclusion: Integration of clinical and radiomic features may enable prediction of postoperative hyponatremia risk. These findings highlight the potential role of radiomic-based risk stratification in postoperative monitoring. A larger, prospective cohort study could help establish radiographic markers for outcomes following pituitary adnenoma resection. Fig. 1 Receive Operating Curve (ROC) showing discrimination of LASSO model for predicting hyponatremia. Fig. 2 Example of MRI input and tumor mask segmentation. Publication History Article published online: 27 February 2026 © 2026. Thieme. All rights reserved. Georg Thieme Verlag KG Oswald-Hesse-Straße 50, 70469 Stuttgart, Germany