The thesis aim is the generation of quantitative biomarkers based on neuroimages for the study of dementia. Hence, through global and structural brain connectivity will be possible to build biomarkers to identify differences between healthy controls and patients. Additionally, with the identification of brain networks altered, would understand how cognition and behavior are altered by these diseases. We have analyzed 160 images from patients with dementia and healthy controls from Hospital Universitarlo San Ignacio, and we have built networks of structural brain connectivity by each patient. All patients had diagnosis by a multidisciplinary group in a memory clinic according to International guidelines of frontotemporal dementiain their variants. To this study we used two types: structutal and diffusion weighted images. With respect to processing we had several had several steps: ¬model estimation, a whole structural connectivity with atlas and statistical analysis. To analyze the differences between variants of frontotemporal dementia and controls, we have made analyze from topological measures from networks and network based statistical. Also, to measure the impasct of this method, we have built a webpage with the main results and asked experts in dementia and neuroradiology about the clinical impact and use of these results. The main results showed the importance of a systematic evaluation with different scales in order to find differences between variant, in other words, the structural connectivity might be confounded by the scale of analysis. The expert raters indicated that the better way to represent connectivity networks were brain maps translucent or circular networks.