Purpose: Positron Emission Tomography (PET) imaging offers the possibility of measuring brain metabolic activity in vivo. However, brain PET images remain difficult to interpret in clinical setting because of the limited spatial resolution of current generation clinical PET scanners. Therefore, the resulting partial volume effect (PVE) is a challenging issue for brain PET image interpretation and quantitative analysis. To overcome this limitation, several algorithms allowing the correction for PVE (PVC) have been developed and assessed mainly in research setting. In this work, we perform a comparative study of 5 different PVC methods using clinical studies. Methods: 17 clinical studies of patients suffering from neurodegenerative disease were included in our study protocol. 3D T1-weighted MRI and FDG-PET were acquired on dedicated MR and PET-CT systems, respectively. MR images were rigidly co-registered to corresponding PET images using the Hermes multimodality platform and segmented using statistical parametric mapping package (SPM8). The resulting images were corrected for PVE using four voxel-based techniques proposed by different groups including Alfano, Muller- Gartner, Meltzer, and Shidahara, and one volume of interest (VOI)-based technique proposed by Rousset. Results: Our results demonstrate a significant increase of the activity concentration in the gray matter. Consequently, the activity in the white matter decreases considerably when using all PVC methods, except for Meltzer and Shidahara. The comparative analysis demonstrates that, among all considered techniques, Alfano's method appears to substantially increase the GM signal. When applying the different PVC methods to specific regions of interest linked to a specific pathology, the results highlight the bias when using uncorrected PET images, but still respecting specific modification patterns of the disease. Conclusion: Our results confirm the necessity of applying PVC to brain PET images in order to obtain more reliable and accurate quantification. This applies particularly to elderly patients with neurodegenerative disease where atrophy induces underestimation of the true PET signal.