Metabolic investigations in patients with hepatic encephalopathy by magnetic resonance imaging and spectroscopy

Hepatic encephalopathy (HE) is a neuropsychiatric disease and a common complication in liver cirrhosis. Several pathogenetic concepts of HE have been proposed in the literature, yet the interplay between those concepts remain a matter of debate. A key factor is attributed to dysfunctional hepatic clearance of neurotoxins in particular, ammonia and its subsequent accumulation in the brain. The neurotransmitter balance in the brain is altered during the detoxification of ammonia, and in consequence, the cognitive, visual, sensorimotor, and eye-hand coordination abilities are hampered. These changes vary from subtle attention and motor deficits to flapping tremor, stupor, and in extreme cases, hepatic coma. To further explore the mechanisms of HE, this dissertation describes improvements of magnetic resonance (MR) imaging and spectroscopy methods to gather metabolic information from brain tissue. In the process, a brain ammonia- or HE-sensitive MR imaging method was established. Thus, the metabolic MR imaging technique chemical exchange saturation transfer (CEST) was optimized for ammonia sensitivity, and subsequently applied in a clinically well-defined cohort of HE patients. Additionally, a simulation framework was designed to investigate limitations in the analysis of γ-aminobutyric acid (GABA)+ due to signal-to-noise ratio (SNR) and linewidth constraints during the acquisition of GABA-edited magnetic resonance spectroscopy (MRS) in deep brain regions. Finally, the neurometabolism in the cerebello-thalamo-cortical pathway of HE patients was investigated with GABA-edited MRS. For the CEST optimization, several model solutions (ammonia, BSA + ammonia & tissue homogenate + ammonia) with varying ammonia concentrations were used to find optimal saturation parameters for an ammonia-weighted contrast. To scrutinize the mechanisms underlying the CEST experiments, protein denaturation and aggregation were assessed in the BSA sample. Finally, the optimal parameters were determined and tested in a healthy control and two patients suffering from HE. In agreement with literature, it was found that amide proton transfer-weighted (APTw) CEST contrast changes in the model solutions were introduced through protein denaturation mediated by ammonia, and an ammonia-sensitive MR imaging contrast was therefore established. In a subsequent study, further phantom measurements revealed ammonia to be the primary driver of the observed contrast changes, while changes associated with other neurometabolites were negligible. In an application study, a cohort of 14 healthy controls, 10 minimal HE and 10 manifest HE patients underwent computer neuropsychometry, critical flicker frequency (CFF) measurements, and blood sample testing prior to the CEST imaging. The processing pipeline of the CEST data included movement correction, magnetization transfer ratio asymmetry (MTRasym) analysis, B0- and B1-correction as well as normalization to minimize possible confounders of the CEST effect. Finally, all MTRasym datasets were normalized to a template (created from the study cohort) to establish pixel-based group statistics over the whole brain volumes. A reduction of APTw signals was found in the cerebellum and the occipital cortex of manifest HE patients. These reduction correlated with blood ammonia levels, as well as several neuropsychometric scores, including motor performance and reaction times. Additionally, increased APTw signals were associated with individual CFF, and thalamic APTw signal alterations were linked to gamma-glutamyltransferase levels. This in vivo study showed patterns and correlations previously found in 13NH3-positron emission tomography (PET) studies, which supports the interpretation that APTw imaging is sensitive to ammonia concentrations. The presented results substantiate literature findings on the involvement of the cerebellum in HE, which is proposed to be involved in motor skill deterioration in HE patients. Additionally, neurotransmitter level alterations, slowed neural oscillations, and decreased visual performance are a common finding in HE, potentially linked to ammonia detoxification in the visual cortex. In the future, the implementation of more advanced CEST techniques may allow to better distinguish underlying contributions - e.g. from CEST mediated by ammonia and contributions from nuclear Overhauser enhancement (NOE) mediated by hepatocerebral degeneration. In vivo MR spectroscopy suffers from low SNR and poor linewidth, resulting in overlapping metabolic resonances. To investigate the impact of spectral quality on data modeling, a nearly noise-free template spectrum was created as the mean of 48 in vivo data sets. Several line broadening and noise level combinations were used to create different scenarios of B0 homogeneity and GABA+ SNR by manipulating template. 100 spectra per condition were quantified with two commonly used MRS quantification tools (Gannet 3.0 and Tarquin). The GABA+ estimation error was quantified as the relative difference between the known modeled GABA+ resonance in the template spectrum, and the modeled GABA resonance of the manipulated spectra. Furthermore, fit error, standard deviation, and the number of rejected spectra were calculated for each combination, and the agreement between the simulations and two in vivo scenarios with varying linewidth and the GABA+ SNR was quantified. The GABA+ estimation error was found to be below 5 % over the entire range of GABA+ SNR for medium linewidth (9.7 Hz) for Gannet 3.0 or all linewidth expect 10.6 Hz for Tarquin. The standard deviation over 100 measurements varied between 3.1 and 17 % for Gannet 3.0 and between 1 and 11 % for Tarquin over the in vivo relevant GABA+ SNR range between 26 and 3.5. The results are in line with other studies on the impact of SNR on the quantification of GABA+. Simulations and in vivo scenarios showed good agreement. The results suggest that GABA-edited studies can be realized for voxels with low GABA+ SNR at the cost of higher group-level variance. Furthermore, B0 homogeneity had negligible effects on the GABA+ modeling. Also, the developed simulation framework could be used to test other quantification approaches. In a final study, the neurometabolism in the cerebello-thalamo-cortical pathway of HE patients was investigated with GABA-edited spectroscopy. The voxels were designed based on the implications drawn from the previous simulation study. The cohort included 16 healthy controls and 16 HE patients (2 minimal HE, 14 HE I), who were assessed via CFF testing, Grooved PEG board testing for motor performance, and blood sample testing. GABA-edited MEscher-GArwood Point RESolved Spectroscopy (MEGA-PRESS) MRS was employed in the cerebellum, the thalamus, and the motor cortex. The GABA-edited difference spectra were analyzed with Gannet 3.0, while the OFF spectra were post-processed with FID-A and quantified with LCModel v6.3. The preliminary results showed increased GABA-to-creatine levels in the cerebellum linked to visual and motor performance scores, which is in line with recent animal and transcranial magnetic stimulation studies. No GABA changes were found in the thalamus and the motor cortex, which contrasts the findings of the animal studies. Further, systemic effects on glutamine, myo-inositol, and aspartate were found in all regions, and were closely linked to visual and motor performance scores. Increased glutamine and myo-inostiol depletion are linked to ammonia detoxification and commonly reported in HE, while changes in aspartate are not reported, yet. Furthermore, increased levels of glutathione, a putative marker of oxidative stress, were found in all regions. The preliminary results indicate an active involvement of the cerebellum in the alterations of the cerebello-thalamo-cortical pathway in HE. The principal results of the present dissertation are the ammonia sensitivity of optimized APTw imaging, and the interpretation of APTw signals as a brain ammonia level correlate. Therefore, APTw imaging could be used to facilitate the understanding of the pathogenesis of HE. Furthermore, the simulation framework for MR spectroscopy could be used during the design and data quality evaluation of GABA-edited MRS studies. Finally, increased levels of GABA in the cerebellum and other systemic effects affecting the neurometabolism in the cerebello-thalamo-cortical pathway were found in HE patients.