Purpose. This study was performed to clarify the role of perfusion-weighted imaging (PWI) at 3 Tesla in the characterisation of haemodynamic heterogeneity within gliomas and surrounding tissues and in the differentiation of high-grade from low-grade gliomas. Materials and methods. We examined 36 patients with histologically verified gliomas (25 with high-grade and I I with low-grade gliomas). PWI was performed by first-pass gadopentetate dimeglumine T2*-weighted echo-planar images, and cerebral blood volume (CBV) maps were computed with a nondiffusible tracer model. Relative CBV (rCBV) was calculated by dividing CBV in pathological areas by that in contralateral white matter. Results. In high-grade gliomas, rCBV were markedly increased in mass [mean standard deviation (SD), 4.3 +/- 1.2] and margins (4.0 +/- 1.1) and reduced in necrotic areas (0.3 +/- 0.3). Oedematous-appearing areas were divided in two groups according to signal intensity on T2-weighted images: tumour with lower (nearly isointense to grey matter) and oedema with higher (scarcely isointense to cerebrospinal fluid) signal intensity. Tumour showed significantly higher rCBV than did oedema (1.8 +/- 0.5 vs. 0.5 +/- 0.2; p < 0.001) areas. low-grade gliomas, mass (2.0 +/- 1.5) and margin (2.2 +/- 1.2) rCBV were significantly lower than in high-grade gliomas (p < 0.001). Conclusions. Three-Tesla PWI helps to distinguish necrosis from tumour mass, infiltrating tumour from oedema and high-grade from low-grade gliomas. It enhances the magnetic resonance (MR) assessment of cerebral gliomas and provides useful information for planning surgical and radiation treatment.