Noninvasive Preoperative Prediction of Transsphenoidal Surgery Outcome in Pituitary GH-Secreting Macroadenomas via Multi-parameter MRI-Based Clini-Radiomic Model: A Multicenter Study.

Authors

Yanghua Fan, Zitong Wu, Wei Liu, Hua Fang, Shuaiwei Guo, Ming Feng, Anna Mou

Published in

Academic radiology. Sep 08, 2026. Epub Sep 08, 2026.

Abstract

Pituitary growth hormone (GH)-secreting macroadenomas pose high surgical difficulty and low transsphenoidal surgery (TSS) remission rates, making preoperative accurate prognosis critical for treatment stratification. Existing predictive models rarely focus on this high-risk subgroup, and though radiomics holds promise for decoding MRI-derived tumor heterogeneity, its use in preoperative TSS outcome prediction for GH macroadenomas is underexplored. Thus, this study aimed to develop and validate a multiparametric MRI-based clini-radiomic model for preoperative prediction of TSS remission in GH-secreting macroadenomas.
In total, 251 patients with GH-secreting macroadenomas were retrospectively divided into training and validation sets. Radiomic features were screened via the elastic net algorithm to construct a fusion radiomic model, whose performance was validated using data from 62 patients at other centers. Subsequently, a clini-radiomic model was developed by integrating this fusion radiomic model with significant clinical features for individualized prediction. Finally, the models' calibration, discrimination, and clinical applicability were comprehensively evaluated using the area under the receiver operating characteristic curve (AUC), accuracy (ACC), calibration curve analysis (CCA), and decision curve analysis (DCA), while SHapley Additive exPlanations (SHAP) was used for feature interpretation.
The fusion radiomic model, built with three key radiomic features, achieved an AUC of 0.930 in the training set, 0.900 in the validation set, and 0.890 in the external validation set. Further integrating this fusion radiomic model with nadir GH level yielded the clini-radiomic model, which exhibited excellent discriminative and calibration performance-with AUCs of 0.960 (training set) and 0.910 (validation set). CCA confirmed good model calibration, while DCA showed net benefits across a wide range of threshold probabilities. SHAP analysis identified nadir GH and radiomic feature 1 as top predictors, with lower values of both correlating with higher remission probability.
The proposed fusion radiomic model shows robust multicenter generalizability, and the integrated clini-radiomic model enables accurate preoperative prediction of TSS outcome in GH-secreting macroadenomas. Its targeted focus on macroadenomas and strict remission criteria address unmet clinical needs, while SHAP-based interpretation enhances transparency. This model can assist clinicians in formulating individualized strategies for high-risk patients.

PMID:
42705923
Bibliographic data and abstract were imported from PubMed on 08 Sep 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.