Comparative analysis of the diagnostic accuracy of FDG PET, single-voxel, and multi-voxel proton MRS for differentiation between radiation-induced necrosis and tumor recurrence was done in 9 patients with brain metastases treated by gamma knife radiosurgery. In all cases enlargement of the lesion and increase of the perilesional edema were demonstrated by MRI on average 10.6+/-2.6 months after initial treatment. Radiation-induced necrosis was identified in 5 patients (histologically in 2, clinically in 3). In one of these a false positive result of FDG PET was observed, whereas data of proton MRS were always correct. The diagnosis of tumor recurrence was established in 4 patients (histologically in 3, clinically in 1). Among these both FDG PET and single-voxel proton MRS showed false negative results (each method twice), whereas multi-voxel proton MRS always permitted us to establish the correct diagnosis. The present study demonstrates the higher diagnostic accuracy of multi-voxel proton MRS, in comparison with single-voxel proton MRS and FDG PET, for the differentiation of the radiation-induced necrosis and tumor recurrence. Its use is especially important in mixed lesions with co-existence of both post-irradiation changes and viable neoplasm. Monitoring of the treatment response by serial multi-voxel proton MRS seems to be reasonable during follow-up of patients with brain metastases after radiosurgery.