Prediction of Pituitary Adenoma's Volumetric Response to Gamma Knife Radiosurgery Using Machine Learning-Supported...

Area of Science:

• Radiology
• Oncology
• Medical Imaging Analysis

Background:

• Gamma Knife radiosurgery (GKRS) is a common treatment for residual or recurrent pituitary adenomas (PA).
• Predicting volumetric tumor response to GKRS is crucial due to observed variability.
• Radiomics, a quantitative imaging analysis, has not been explored for predicting PA response to GKRS.

Purpose of the Study:

• To pioneer the use of radiomic MRI analysis to predict the volumetric response of pituitary adenomas to GKRS.
• To develop and evaluate radiomics models for predicting treatment outcomes in PA patients undergoing GKRS.

Main Methods:

• Retrospective observational cohort study of 81 patients treated with GKRS for PA.
• Extraction of radiomic features (intensity, shape, texture) from pre-treatment 3-Tesla MRI scans.
• Application of LASSO for feature selection and various classifiers (random forest, SVM, etc.) to build predictive models.

Main Results:

• Radiomics models achieved high predictive performance with AUC values up to 0.928 and R² up to 0.665.
• Single-sequence (T1w) and dual-sequence (T1w + CE-T1w) radiomics models showed strong prognostic capabilities.
• Multi-modality models incorporating clinicopathological (CP) parameters (CP + T1w + CE-T1w) also performed well (AUC 0.909).
• All developed radiomics models significantly outperformed a benchmark model using only CP features (AUC 0.846).

Conclusions:

• This study demonstrates the potential of MRI-based radiomics for predicting volumetric response in pituitary adenomas treated with GKRS.
• Radiomics models offer superior predictive performance compared to traditional clinicopathological parameters.
• These findings support the use of radiomics for treatment individualization in PA management.