Magnetic resonance imaging (MRI) is increasingly being used in medical settings because of its ability to produce, non-invasively, high quality images of the inside of the human body. Since its introduction in early 70’s, more and more complex acquisition techniques have been proposed, raising MRI to be exploited in a wide spectrum of applications. Innovative MRI modalities, such as diffu- sion and functional imaging, require complex analysis techniques and advanced algorithms in order to extract useful information from the acquired data. The aim of the present work has been to develop and optimize state-of-the- art techniques to be applied in the analysis of MRI data both in experimental and clinical settings. During my doctoral program I have been actively involved in several research projects, each time facing many different issues. In this dissertation, however, I will report the results obtained in three most appealing projects I partecipated to. These projects were devoted (i) to the implementation of an innovative experimental protocol for functional MRI in laboratory animals, (ii) to the development of new methods for the analysis of Dynamic Contrast Enhanced MRI data in experimental tumour models and (iii) to the analysis of diffusion MRI data in stroke patients. Particular emphasis will be given to the technical aspects regarding the algorithms and processing methods used in the analysis of data. Apart from a brief introduction on magnetic resonance imaging principles, the dissertation is organised in three parts, each one of them covering a separate topic dealt with during my doctoral studies. Each chapter is self-contained, in that it gives an introduction about the issue faced in the study, describes the methods and the techniques employed, and concludes with a discussion about the results obtained. Chapter 3 illustrates the methodology proposed by our group for an innovative method of fMRI (Activation-Induced Manganese Enhanced MRI, AIM-MRI) in rats. The method is based on in-vivo, absolute quantification of Mn concentration in different brain regions performed by fast T1 mapping and coregistration to a rat brain atlas. This strategy allows to quantify the accumulation of Mn in different regions of the rat brain, which is strictly related to the ac- tivation status of each area. This mechanism is similar to what happens in classical functional MRI studies but, in addition, it might be used for functional experiments performed in awake animals. In chapter 4 we cover the topic of tissues classification from MRI images in some experimental tumour models and present some advanced image-processing techniques we introduced for the analysis of Dynamic Contrast-Enhanced MRI (DCEMRI) data. In particular, we proposed to estimate a set of characteristic features from DCEMRI time profiles which well describe their shape, and then to extract peculiar behaviours which discriminate each tissue inside the tumour by combining cluster analysis and machine learning techniques. We have tested several approaches, each time comparing them with the state-of-the-art techniques of analysis for this kind of data. Finally, the proposed approach has been validated in a real case application in order to assess the efficacy of an anti-cancer therapy. Chapter 5 addresses the study on post-stroke plasticity carried out in collab- oration with the Signal Processing Laboratory of the Swiss Federal Institute of Technology in Lausanne (EPFL). Brain structural connectivity was monitored with Diffusion Spectrum Imaging (DSI), i.e. a high angular resolution diffusion MRI technique, during the functional recovery from the stroke, at hyperacute, acute and chronic stages after the lesion onset. The reproducibility of extracted fibre tracts has been extensively studied on healthy subjects, and the parameter set of fiber-tracking algorithm giving the best results was then used in patient analysis. The extracted fibre bundles between each pair of cortical regions were characterized by means of several connectivity measures.