Abstract
BACKGROUND: Brain metastases (BM) are increasingly treated with stereotactic radiosurgery (SRS); however, up to 30% of BM recur locally. This work investigated whether hyperpolarized (HP) [1-(13)C]-pyruvate MRI can be used to predict SRS treatment response in patients with BM. METHODS: Eighteen patients with 44 BM were imaged with HP [1-(13)C]-pyruvate MRI prior to SRS. Treatment response was determined using the Response Assessment in Neuro-Oncology BM (RANO-BM) working group guidelines at 6-month follow-up. Fourteen parameters, including lesion [1-(13)C]-lactate to [(13)C]-bicarbonate, [1-(13)C]-lactate to [1-(13)C]-pyruvate and [(13)C]-bicarbonate to [1-(13)C]-pyruvate signal ratios, in addition to prognostic and dosimetric parameters, were analyzed using univariable and multivariable analysis. RESULTS: Univariable analysis identified lesion [1-(13)C]-lactate to [(13)C]-bicarbonate ratio (P = .0003), lesion [(13)C]-bicarbonate to [1-(13)C]-pyruvate ratio (P = .0118), lesion volume (P = .0264), and the number of involved organs with metastases including the brain (P = .0448) as significant predictors of treatment response. The lesion [1-(13)C]-lactate to [(13)C]-bicarbonate ratio was predictive of response with the best overall performance, achieving an AUC(ROC) = 0.88, AUC(PRC) = 0.83, sensitivity = 67% (CI: 40%-87%), specificity = 97% (CI: 90%-100%), and positive predictive value (PPV) = 91% (CI: 73%-100%). CONCLUSIONS: HP lesion [1-(13)C]-lactate to [(13)C]-bicarbonate ratio can predict SRS response with a high PPV.