In most melancholic patients, melancholia may be considered as a biologically homogeneous clinical entity, especially when compared with other depression subtypes. In addition to such features, a number of studies have also SB431542 in vivo identified brain structural alterations in melancholia, such as volume reductions in the hippocampus, right anterior supplementary motor area or left insula. Interestingly, in a subset of these studies, alterations in cerebro-spinal fluid spaces were also described. Furthermore, in one of the studies, CSF increases in the left Sylvian fissure were related to the time to remission of the depressive episode, thus giving clinical relevance to the findings. Voxel-wise methods such as voxel based morphometry are particularly appropriate for the study of brain structural alterations as they offer an unbiased estimation of whole-brain abnormalities. However, since VBM strongly depends on accurate brain tissue segmentation, CSF alterations have rarely been assessed using such procedures as CSF segmentation has traditionally been regarded as a rather unreliable approach. Indeed, although in most segmentation algorithms CSF is accurately isolated from gray and white matter, it is not uncommon for segmented CSF images to include voxels from non-brain structures such as the dura, the venous sinuses, the scalp or the skull. Consequently, until now we have not attempted to directly replicate with a voxel-wise technique the CSF findings previously described in melancholic samples using ROI approaches. Nevertheless, the development of new tissue segmentation algorithms, such as the so-called ‘new segment’ algorithm, may help to overcome such limitations, as it provides further information as to the a priori distribution of nonbrain tissue. The new algorithm should prevent misclassification of non-brain voxels as CSF. The aim of this study was to assess whole-brain voxel-wise alterations in the CSF spaces of a sample of melancholic patients in comparison to a group of control subjects of similar age and gender distribution. To evaluate the benefits of using the ‘new segment’ algorithm, we compared the results obtained using such a method with those obtained by means of the ‘unified segmentation’ approach, as implemented in both SPM5 and SPM8. In addition, the most relevant between-group differences observed with the ‘new segment’ algorithm were validated by means of a non-automated ROI analysis. Finally, to evaluate the relevance of CSF alterations, findings were correlated with the clinical data of the study sample. To our knowledge, this is the first whole-brain voxel-wise study comparing CSF volumes of melancholic patients with healthy controls. Specifically, we observed a significant CSF increase in the region of the left Sylvian fissure and a CSF volume decrease at the level of the medial and lateral parietal cortex. We also observed that such findings were not detected using earlier segmentation algorithms.