Predictive value of early DCE and DSC perfusion MRI parameters for midterm clinical outcomes in lung cancer brain metastases treated with stereotactic radiosurgery.

PURPOSE

Stereotactic Radiosurgery (SRS) is an effective way of controlling the brain metastasis (BM) of lung carcinoma. This study evaluates the performance of dynamic contrast-enhanced MRI (DCE-MRI) and dynamic susceptibility contrast MRI (DSC-MRI) parameters to distinguish responders from non-responders at midterm follow-up in lung carcinoma BMs.

METHODS

Twenty-six patients (mean age 62 ± 10 years) with 54 lung carcinoma BMs (NSCLC 67%, SCLC 33%) underwent SRS. The DCE-MRI and DSC-MRI were performed at baseline and repeated 4-8 weeks post-SRS to predict treatment responses at the midterm follow-up (6-12 months). Midterm outcomes were classified according to RANO-BM criteria as responders (complete response, partial response, or stable disease) or non-responders (progressive disease). Receiver operating characteristic (ROC) analyses evaluated the diagnostic accuracy of individual perfusion parameters and their combinations.

RESULTS

Forty lesions (74%) were responders, while 14 (26%) were non-responders. Progressive lesions showed a mean volume increase of 5.5-fold, whereas responders demonstrated a 60% mean volume reduction. Responders showed significantly lower absolute post-SRS K-trans (median 0.015 vs. 0.035 min⁻¹; p = 0.005), a higher proportional decrease in K-trans from baseline (- 27% vs. +13%; p = 0.017), and lower post-SRS Ve (p = 0.009) compared to non-responders. Absolute post-SRS K-trans had the highest individual predictive accuracy (AUC = 0.75, accuracy = 78%, sensitivity = 86%, specificity = 55%). Neither the dynamic change nor post-SRS nCBV alone predicted midterm response; however, combining post-SRS nCBV with K-trans slightly improved predictive performance (AUC = 0.76, accuracy = 79%).

CONCLUSION

Early post-SRS absolute K-trans is the best perfusion parameter for predicting midterm response in lung carcinoma BMs. DSC-MRI parameters alone offer limited predictive value.