MRI-based quantification of brain damage in cerebrovascular disorders

Abstract

Brain diseases can lead to diverse structural abnormalities that can be assessed on magnetic resonance imaging (MRI) scans. These abnormalities can be quantified by (semi-)automated techniques. The studies described in this thesis aimed to optimize and apply cerebral quantification techniques in patients with cerebrovascular disorders. The first part of this thesis compared quantification techniques for brain volume change measurements (Unified segmentation (US), k-nearest neighbor-based probabilistic segmentation (kNN) and structural image evaluation, using normalization, of atrophy (SIENA)). We showed that US and kNN have a good precision, accuracy and comparability for cross-sectional brain volume measurements. For measurements of volume change, SIENA showed the best performance, but kNN is a good alternative if volume change measurements of other brain structures are required. We also compared different image processing conditions for the detection of cerebral microbleeds on 7T MRI. We showed that 7 Tesla MRI has a high sensitivity for microbleed detection. However, even with optimal image processing the limited size of some of the lesions and susceptibility effects of adjacent other structures complicated visual detection, which led to a modest inter-rater agreement of visual rating on 7 Tesla MRI. Automated lesion detection techniques may be required to optimally benefit from the increased spatial resolution offered by 7 Tesla MRI. The second part of this thesis assessed the effects of type 2 diabetes mellitus (T2DM) on the brain by neuropsychological evaluation and quantitative brain volume measurements. In a longitudinal study we observed that there was a greater increase in lateral ventricular volume over time in patients with T2DM as compared to controls, indicative of a modest increase in the rate of brain atrophy over the course of years. We also observed that albuminuria predicted accelerated cognitive decline in patients with T2DM, but other microvascular complications were unrelated to accelerated cognitive decline or brain MRI abnormalities. In cross-sectional analyses we observed that the effects of T2DM on cortical grey matter were most pronounced in the medial temporal lobe, although the volume of the hippocampus, a medial temporal lobe structure relevant to learning and memory, was not affected. The third part of this thesis assessed the effects of aneurysmal subarachnoid hemorrhage (aSAH) on the brain by quantitative brain volume measurements. The kNN procedure was optimized and validated for segmentation of brain volumes on MRI in patients after aSAH. Patients 6 months after aSAH who had a relatively good outcome still had larger ventricles and smaller peripheral cerebrospinal fluid volumes than controls. Worse functional outcome in patients was related to reduced parenchymal and increased ventricular volumes. Brain volumes still changed from 6 to 18 months after aSAH. Condition on admission, complications and outcome were related to brain volumes 6 and 18 months after aSAH. In conclusion, we optimized cerebral quantification techniques to assess brain abnormalities in patients with cerebrovascular disorders. The added value of these techniques was demonstrated in etiological studies on patients with T2DM and patients after aSAH. These techniques are of value for future etiological and intervention studies.