Pituitary adenomas (PAs) are among the most prevalent lesions of the sella turcica, accounting for 10%-25% of all intracranial neoplasms. Pituitary macroadenomas (PMAs) are defined with a maximum diameter of over 1 cm. Tumor characteristics are key factors influencing surgical effectiveness and complications of PMAs, with tumor perfusion and consistency identified as major predictive factors in literature. Conventional sequences provide limited information for predicting the perfusion and consistency of pituitary adenomas. Advanced sequences offer additional insights. However, the efficacy of combining radiomic features from multiparametric sequences, incorporating both conventional and advanced sequences, has not yet been proved.

We aim to develop machine learning models that combines radiomic features developed from both conventional and advanced sequences to predict the perfusion and consistency of PMAs. Furthermore, we aim to demonstrate the clinically applicability of these models by constructing a MR-PIT stratification (Multiparametric Radiomic derived and tumor Perfusion and consIsTency based surgical difficulty stratification), which correlated with the surgical strategy and outcomes.

Inclusion Criteria:

• patients with tumor more than 2.5 cm of maximal diameter in the coronal plane
• Functional and non-functional pituitary tumors

Exclusion Criteria:

• incomplete image data

Patients histologically diagnosed with pituitary adenomas and who had undergone surgical treatment