Predictive Modeling of Non-functioning Giant Pituitary Neuroendocrine Tumor Resection: A Multi-Planar Perspective.

Background: Giant pituitary neuroendocrine tumor (GPitNET) are challenging tumors with low rates of gross total resection (GTR) and high morbidity. Previously reported machine-learning (ML) models for prediction of pituitary neuroendocrine tumor extent of resection (EOR) using preoperative imaging included a heterogenous dataset of functional and non-functional pituitary neuroendocrine tumors of various sizes leading to variability in results.

Objective: The aim of this pilot study is to construct a ML model based on the multi-dimensional geometry of tumor to accurately predict the EOR of non-functioning GPitNET.

Methods: A retrospective study of 100 large non-functioning GPitNETs (≥3 cm diameter, >10 cm³ volume) was conducted to develop predictive models for GTR or EOR based on 5 variables: tumor diameter, shape, revised Knosp grade, and modified Hardy classifications for sellar and extrasellar invasion. Model performance was assessed using ROC-AUC and confusion matrix metrics.

Results: The median pre-operative tumor volume was 17.35 cm(IQR: 12.4-27.0).The median EOR was 97.6% (IQR: 84.9-100), and GTR was achieved in 49% of patients. The most predictive variables were the modified Hardy classification for extrasellar extension and Knosp Grade (AUC of 0.771 and 0.713,respectively). Among the constructed ML models, the XGBoost algorithm had the highest predictive capability, with an internal validation AUC of 0.86,while the external validation sensitivity, specificity, positive, and negative predictive values were 84%, 77%, 78% and 82%, respectively.

Conclusion: Utilizing preoperative imaging parameters in a 3-dimensional manner proves highly valuable in predicting the EOR for non-functioning GPitNETs.These predictions can be easily calculated using an online open-access application: https://anonymousforreview.shinyapps.io/predicted_giant_pituitary_adenoma_resection/.