Grading of cerebral glioma with multiparametric MR imaging and 18F-FDG-PET: concordance and accuracy.

Abstract

To retrospectively evaluate concordance rates and predictive values in concordant cases among multiparametric MR techniques and FDG-PET to grade cerebral gliomas. Multiparametric MR imaging and FDG-PET were performed in 60 consecutive patients with cerebral gliomas (12 low-grade and 48 high-grade gliomas). As the dichotomic variables, conventional MRI, minimum apparent diffusion coefficient in diffusion-weighted imaging, maximum relative cerebral blood volume ratio in perfusion-weighted imaging, choline/creatine ratio and (lipid and lactate)/creatine ratio in MR spectroscopy, and maximum standardised uptake value ratio in FDG-PET in low- and high-grade gliomas were compared. Their concordance rates and positive/negative predictive values (PPV/NPV) in concordant cases were obtained for the various combinations of multiparametric MR techniques and FDG-PET. There were significant differences between low- and high-grade gliomas in all techniques. Combinations of two, three, four, and five out of the five techniques showed concordance rates of 77.0 ± 4.8%, 65.5 ± 4.0%, 58.3 ± 2.6% and 53.3%, PPV in high-grade concordant cases of 97.3 ± 1.7%, 99.1 ± 1.4%, 100.0 ± 0% and 100.0% and NPV in low-grade concordant cases of 70.2 ± 7.5%, 78.0 ± 6.0%, 80.3 ± 3.4% and 80.0%, respectively. Multiparametric MR techniques and FDG-PET have a concordant tendency in a two-tiered classification for the grading of cerebral glioma. If at least two examinations concordantly indicated high-grade gliomas, the PPV was about 95%. • Modern imaging techniques can help predict the aggressiveness of cerebral gliomas. • Multiparametric MRI and FDG-PET have a concordant tendency to grade cerebral gliomas. • Their high-grade concordant cases revealed at least 95 % positive predictive values. • Their low-grade concordant cases revealed about 70–80% negative predictive values.