Predicting local control of brain metastases after stereotactic radiosurgery with clinical, radiomics and deep learning features

Abstract Background and purpose Timely identification of local failure after stereotactic radiotherapy for brain metastases allows for treatment modifications, potentially improving outcomes. Previous studies showed that adding radiomics or Deep Learning (DL) features to clinical features increased Local Control (LC) prediction accuracy. However, no study has integrated radiomics, DL, and clinical features into machine learning algorithms to predict LC. We examined whether a model using all these features achieves better accuracy than models using only a subset. Materials and methods We collected pre-treatment brain MRIs and clinical data for 129 patients at the Gamma Knife Center of the Elisabeth-TweeSteden Hospital. Radiomics features (extracted using the Python radiomics feature extractor) and DL features (extracted using a 3D ResNet model) were combined with clinical features. Performance of a Random Forest classifier was compared across four models trained with: clinical features only; clinical and radiomics features; clinical and DL features; and clinical, radiomics, and DL features. Results The prediction model utilizing only clinical variables provided an Area Under the receiver operating characteristic Curve (AUC) of 0.82 and an accuracy of 75.6%. Adding radiomics features increased the AUC to 0.88 and accuracy to 83.3%, while adding DL features resulted in an AUC of 0.86 and accuracy of 78.3%. The best performance came from combining clinical, radiomics, and DL features, achieving an AUC of 0.89 and accuracy of 87%. Conclusion Integrating radiomics and DL features with clinical characteristics improves LC prediction after stereotactic radiotherapy for brain metastases. These findings demonstrate the potential for early outcome prediction, enabling timely treatment modifications to improve patient management. Clinical and Translational Impact Statement Our study holds great clinical value, as the increased prediction accuracy can lead to tailored and effective interventions, resulting in better outcomes for brain metastases patients treated with stereotactic radiotherapy.