AB084. SOH25_AB_198. Radiomics in pituitary tumours: a systematic review and meta-analysis of diagnostic, prognostic and classification efficacy

Background: Pituitary tumours are diverse sellar neoplasms classified by functionality, molecular characteristics and vascularity. Accurate subclassification is crucial for treatment and prognosis, however traditional imaging often misses subtle heterogeneity. Radiomics extracts quantitative features from medical images, revealing tumour phenotypes beyond visual assessment, which aids sub-classification. The application of radiomics to pituitary tumours is emerging. This review aims to evaluate radiomic features in pituitary tumour sub-classification, prognostication and textural analysis and to determine methodologies in radiomic analysis.

Methods: A comprehensive literature search was conducted in MEDLINE, Web-of-Science and Scopus, focusing on peer-reviewed articles in English. Inclusion criteria included studies on radiomic features for pituitary tumours with details on sub-classification, prognostication or textural analysis, and methodologies such as imaging, preprocessing, feature extraction, selection and statistical analysis. The data were synthesised qualitatively and quantitatively, with meta-analysis for comparable studies.

Results: A total of 182 articles were identified, 76 were screened and 47 studies met the inclusion criteria. The review highlights several key findings. Studies demonstrated that magnetic resonance imaging (MRI)-based texture features could distinguish craniopharyngiomas from meningiomas, predict non-functioning pituitary tumour consistency and identify high risk adenomas. Radiomics also showed promise in predicting the Ki-67 index, postoperative visual recovery and preoperative tumour subtypes. Integrating radiomic features with clinical data consistently improved predictive model performance, highlighting the potential for enhanced clinical decision-making in pituitary tumour management.

Conclusions: Radiomics significantly enhances the diagnostic and prognostic capabilities for pituitary tumours through advanced MRI-based texture analysis and machine learning. Integrating radiomic features with clinical data offers promising improvements in clinical decision-making and patient outcomes.

Keywords: Machine learning; neurosurgery; pituitary; radiomics; tumours