Thalamus is a structure in the brain that controls states of awakeness, sleeping and consciousness. Neurodegenerative diseases like multiple sclerosis (MS) cause damage to nerve cells including cells in the thalamus. With segmentation of thalamus in magnetic resonance (MR) images we could calculate volume of thalamus and relate it to the disability caused by the disease. In this thesis we developed an automated segmentation method based on co-registration of atlases, which contain thalamus segmentations, onto the target MR image and fusion of the atlases. The dataset of atlases was created by manually segmenting the MR images. Automated segmentation of the thalamus with affine and B-spline based registration and majority voting atlas fusion yielded best quality according to quantitative evaluation. The evaluation was based on T1-weighted MR images of 87 subjects with MS and 88 healthy subjects. We also investigated the relationships between the (normalized) thalamic volumes and the clinical and demographic data and compared the volume between MS and healthy subjects. Similarly to previous reports in the literature the thalamic volumes decreased with increasing age, extended disease duration, and increasing functional disability of the MS patients. On the other hand, the thalamic volumes increased with years of education and results of the cognitive memory tests. By measuring the volume of thalamus we could thus enable higher quality of life of the MS patients as a result of the possibility of earlier and more potent treatment.