Conversely, however, the parietal CSF volume reductions reported here are unlikely to be of neurodegenerative origin. Nevertheless, if we consider depression as a chronic and recurrent disease with multiple genetic and environmental determinants, complex models are plausible, and thus structural abnormalities of neurodegenerative origin may well coexist with alterations of another etiology across different brain areas over the course of the illness. In addition to the above ideas, the CSF alterations reported here seem to be of clinical relevance as, in our sample of melancholic patients, larger CSF volumes in the left Sylvian fissure were related to a smaller improvement of depressive symptomatology at discharge. This observation is in accordance with previous correlations obtained with closely related measures, such as regional CSF insular volume or gray matter content of the insular cortex, which were related to the time to remission of the depressive episode after treatment initiation. Indeed, we also observed a correlation between CSF volume increase and time to remission, although the result was no longer significant after Bonferroni correction for multiple comparisons. Furthermore, there is also functional evidence supporting the relationship between insular cortex and clinical outcome, such as the observations of a normalization of insular hyperactivity after clinical recovery or successful antidepressant treatment. Contrasting the results obtained with the two different image segmentation strategies used here, we observed that the CSF content of the left Sylvian fissure was only significantly increased when data were pre-processed using the ‘new segment’ algorithm. Although there is no ‘gold-standard’ for comparing the results obtained with the two pre-processing approaches, we validated the left Sylvian fissure CSF volume increase detected here under the ‘new segment’ approach with a non-automated ROI analysis. Left, but not right, Sylvian fissure showed a lower mean signal intensity in combination with a larger volume. Also, such a finding is in accordance with earlier studies reporting brain structural alterations in melancholia also using non-automated image analysis techniques. Furthermore, the same CSF increase was also detected using the ‘unified segmentation’ algorithm, Doxorubicin albeit at a non-significant level, suggesting a reduced sensitivity of the latter approach to detect regional changes in CSF volume. Parietal CSF volume decreases were also uniquely detected using the ‘new segment’ approach, although, in this case, we did not attempt to replicate the findings with a non-automated ROI approach since findings were located in less definite regions in anatomical terms. Global CSF measurements also differed between the two segmentation methods studied.