Background MRI radiomic features and machine learning have been used to predict brain metastasis (BM) stereotactic radiosurgery (SRS) outcomes. Previous studies used only single centre datasets, representing a significant barrier to clinical translation and further research. This study therefore presents the first dual-centre validation of these techniques. Methods SRS datasets were acquired from two centres (n=123 BMs and n=117 BMs). Each dataset contained eight clinical features, 107 pre-treatment T1w contrast-enhanced MRI radiomic features, and post-SRS BM progression endpoints determined from follow-up MRI. Random decision forest models were used with clinical and/or radiomic features to predict progression. 250 bootstrap repetitions were used for single centre experiments. Results Training a model with one centre’s dataset and testing it with the other centre’s dataset required using a set of features important for outcome prediction at both centres, and achieved area under the receiver operating characteristic curve (AUC) values up to 0.70. A model training methodology developed using the first centre’s dataset was locked and externally validated with the second centre’s dataset, achieving a bootstrap-corrected AUC of 0.80. Lastly, models trained on pooled data from both centres offered balanced accuracy across centres with an overall bootstrap-corrected AUC of 0.78. Conclusions Using the presented validated methodology, radiomic models trained at a single centre can be used externally, though they must utilize features important across all centres. These models’ accuracies are inferior to those of models trained using each individual centre’s data. Pooling data across centres shows accurate and balanced performance, though further validation is required.