To date only post-mortem evidence exists that hypothalamic dysregulations in affective disorders might be associated with hypothalamic volume changes. Sub-millimeter resolution provided by 7T MRI now allows us to investigate these changes in-vivo.T1-weighted brain images from 20 unmedicated and 20 medicated patients with major depression, 21 patients with bipolar disorder, and 23 healthy controls were acquired with the MP2RAGE sequence and a Magnetom 7T (Siemens). The images were intensity standardised to a representative sample image using a global linear scaling factor estimated from two subcortical ROIs of WM and GM. A detailed algorithm developed for the segmentation of the human hypothalamus on 7T T1-weighted MR images was applied. It employs a specialised colour-coding to ensure optimal reliability. Three trained raters, who were blind for diagnosis and brain hemisphere, delineated the hypothalamus on the randomised images (inter-rater reliability: ICC=0.89; 20 hypothalami unilateral).There was a hemispheric asymmetry for the hypothalamus volume by trend (right>left; t(83)=−1.92, p=0.059), a significant gender effect for both hypothalami (men>women; left: t(82)=4.56, p<0.001; right: t(82)=4.49, p<0.001), and a significant correlation with the intracranial volume (ICV) (left: r=0.49, p<0.001; right: r=0.56, p<0.001). Analyses of covariance, controlling for ICV and sex revealed that the main factor group explained a significant portion of variance of the left hypothalamus volume (F(3)=2.83, p=0.044, ηp2=0.10) but not of the right hypothalamus volume (F(3)=1.30, p=0.28, ηp2=0.05). A simple contrast showed that the left hypothalamus was enlarged by trend in the unmedicated depressive patients (p=0.052) and significantly enlarged in both medicated patient groups (major depression: p=0.008; bipolar disorder: p=0.046).Our in-vivo findings stand in contrast to post-mortem volume reductions of the hypothalamus but they are in line with post-mortem volume increases of hypothalamic nuclei. The role of sample characteristics, methodology, and covariates will be discussed.