Combining Anatomical and Spectral Information to Enhance MRSI Resolution and Quantification: Application to Multiple Sclerosis

Multiple sclerosis is a progressive autoimmune disease that affects young adults. Magnetic resonance (MR) imaging has become an integral part in monitoring multiple sclerosis disease. Conventional MR imaging sequences such as fluid attenuated inversion recovery imaging have high spatial resolution, and can visualise the presence of focal white matter brain lesions in multiple sclerosis disease. Manual delineation of these lesions on conventional MR images is time consuming and suffers from intra and inter-rater variability. Among the advanced MR imaging techniques, MR spectroscopic imaging can offer complementary information on lesion characterisation compared to conventional MR images. However, MR spectroscopic images have low spatial resolution. Therefore, the aim of this thesis is to automatically segment multiple sclerosis lesions on conventional MR images and use the information from high-resolution conventional MR images to enhance the resolution of MR spectroscopic images. Automatic single time point lesion segmentation is performed using T1-weighted and fluid attenuated inversion recovery MR images where the brain is segmented into grey matter, white matter, cerebrospinal fluid and lesions using a probabilistic framework optimised by the expectation-maximisation technique. Then, a patch-based super-resolution method is used to increase the spatial resolution of metabolite maps computed from MR spectroscopic imaging. The patch-based super-resolution method uses high-resolution T1-weighted and fluid attenuated inversion recovery images together with the brain segmentations to regularise the super-resolution process. Finally, we extend the single time point lesion segmentation idea to two time points lesion segmentation. The two time points lesion segmentation is performed using T1-weighted and fluid attenuated inversion recovery MR images of the two time points optimised using a joint expectation-maximisation algorithm. Validation of lesion segmentation methods on clinical datasets shows that they are accurate and consistent in segmenting multiple sclerosis lesions. Moreover, we study the association of a clinical biomarker -- the Expanded Disability Status Scale score -- and a MR imaging biomarker -- new and enlarging lesion volume -- and we observe that in majority of the patients, the new and enlarging lesion volume correlates better with Expanded Disability Status Scale score's evolution at patient level than at a group level. This study also shows the applicability of our method on multi-centre datasets without re-tuning or re-training, which makes it useful for the clinical use. Validation of the super-resolution method on synthetic and real images shows that our method preserves tissue contrast and structural information; and matches well with the trend of acquired high-resolution MR spectroscopic images. We analyse N-acetyl aspartate and myo-Inositol metabolites concentration in lesions and in the surrounding white matter. N-acetyl aspartate metabolite concentration in lesions is found to be lower compared to the surrounding white matter, and an opposite trend is observed for the myo-Inositol metabolite concentration. From this research we conclude that the developed multiple sclerosis lesion segmentation methods, through their robustness and automation, could bring an added value to the clinical routine evaluation of multiple sclerosis patients. Also, the patch-based high-resolution MR spectroscopic images, through its tissue contrast conservation, can offer better lesion characterisation.status: publishe