Imaging features of recurrent brain metastasis (rBM) and radiation necrosis (RN) (Aim 1)

Will compute the diffusion-relaxation (DR) spectra and total sodium concentration (TSC) within biopsy regions of interest and use Student t-test or Mann-Whitney U test, depending on the data distribution, to compare imaging features between rBM and RN groups. Specifically, for DR spectra, the test statistic at each spectral location will be used to identify spectral components associated viable tumor and treatment effect. The information will subsequently be used to partition the voxel-wise DR spectra and compute voxel-wise "viable tumor" and "treatment effect" fraction maps. Multivariate logistic regression models will assess the predictive value of the imaging biomarkers, including DR-correlation spectrum imaging (CSI) spectral components and TSC. Receiver operating characteristic (ROC) analysis will evaluate diagnostic performance, calculating area under the curve, sensitivity, specificity, and optimal cutoff values for differentiating rBM and RN.

At time of advanced magnetic resonance imaging: Baseline

Stereotacic imaging-guided biopsies and correlation with tissue quantifications (Aim 1)

Will prospectively acquire DR-CSI and sodium imaging-guided biopsies of tumor tissue in regions with different imaging characteristics. Pearson or Spearman correlations will be computed between imaging features and quantitative histological measures of cellularity, extracellular fraction, and percentages of tumor and necrosis.

Up to completion of surgery

Early imaging changes as predictors of clinical outcomes (Aim 2)

Will compute DR-CSI spectral components and TSC at each time point. Repeated measure analysis of variance will assess imaging changes over time. Will also plot trajectories of imaging biomarkers to visualize temporal patterns. Multivariate logistic regression models will evaluate early imaging changes (2 weeks and 3 months) as predictors of 6-month outcomes. Lesions will be stratified into responders (complete or partial response) and non-responders (stable disease or progression) based on their Response Assessment in Neuro-Oncology Brain Metastases assessment at 6-month post-SRS. ROC curves will assess the predictive performance of early microstructural imaging changes and the multivariate model. Primary tumor type will be included as a covariate in analyses.

At baseline and 2 weeks, 3 months, and 6 months post-stereotactic radiosurgery (SRS)