Met de vergrijzing van de bevolking groeit ook het belang van een vroege en nauwkeurige diagnose van neurodegeneratieve aandoeningen zoals de ziekte van Alzheimer (AD). Dit onderzoek richt zich...Show moreMet de vergrijzing van de bevolking groeit ook het belang van een vroege en nauwkeurige diagnose van neurodegeneratieve aandoeningen zoals de ziekte van Alzheimer (AD). Dit onderzoek richt zich specifiek op het verkennen van de verschillen tussen twee vormen van AD, namelijk Early-Onset Alzheimer's Disease (EOAD) en Late-Onset Alzheimer's Disease (LOAD), met als doel het verbeteren van classificatiemodellen. We hebben vier classificatiemodellen ontworpen, elk gericht op specifieke hersengebieden die mogelijk atrofie vertonen, met Model 1 en 2 gericht op respectievelijk LOAD en EOAD, Model 3 als een gecombineerd model en Model 4 dat de algehele hersenatrofie omvat. Deze modellen zijn geanalyseerd aan de hand van Area Under the Curve (AUC) en zijn ontwikkeld met behulp van logistische regressie en LASSO om de meest relevante variabelen te selecteren. Onze bevindingen suggereren dat EOAD en LOAD verschillende neuropathologische patronen vertonen, waarbij EOAD mogelijk wordt gekenmerkt door specifiekere neuropathogenese patronen, terwijl LOAD meer uniforme hersenveranderingen vertoont. Model 3, dat specifieke hersengebieden combineerde, presteerde het beste bij het classificeren van beide groepen, terwijl Model 4, dat alle hersengebieden omvatte, een lagere classificatie-accuratesse vertoonde voor EOAD. Deze resultaten benadrukken dat AD patiënten geen homogene groep zijn en benadrukken het belang van gedifferentieerde diagnostische benaderingen om rekening te houden met de heterogeniteit binnen deze aandoening.Show less
Achtergrond: Anatomische MRI blijkt een goede voorspeller te zijn van Alzheimer, het is echter nog niet heel duidelijk wat de rol van functionele MRI (fMRI) in Alzheimer classificatie is en of dit...Show moreAchtergrond: Anatomische MRI blijkt een goede voorspeller te zijn van Alzheimer, het is echter nog niet heel duidelijk wat de rol van functionele MRI (fMRI) in Alzheimer classificatie is en of dit een toevoeging op anatomische MRI kan zijn. Doel: In dit onderzoek wordt onderzocht of fMRI van toegevoegde waarde is op anatomische MRI in het classificeren van Alzheimer. Methode: Om dit te kunnen onderzoeken worden er 3 classificatiemodellen met elkaar vergeleken, namelijk een model met alleen anatomische MRI-predictoren bestaande uit de grijze stofdichtheid en subcorticale volumes en een model met alleen fMRI-predictoren bestaande uit de functionele connectiviteit tussen 20 hersengebieden; en ten slotte een multimodaal model met zowel de anatomische als fMRI-predictoren. Om de meest relevante predictoren in de modellen te selecteren wordt er een LASSO algoritme toegepast. Daarnaast wordt er kruisvalidatie toegepast om de modellen te kunnen generaliseren. Om de prestatie van de modellen te evalueren worden de AUC-waardes van de modellen vergeleken. Resultaten: Het model dat alleen gebruikmaakt van anatomische MRI-predictoren kan Alzheimer accuraat classificeren met een AUC-waarde van 0.92. Het model dat alleen fMRIpredictoren gebruikt kan Alzheimer redelijk goed classificeren, maar niet beter dan model 1, met een AUC-waarde van 0.74. Het multimodale model classificeert Alzheimer het beste met een AUC-waarde van 0.93. Conclusie: Een multimodaal classificatiemodel bestaande uit anatomische MRI en fMRI presteert beter dan een model met alleen één van deze modaliteiten. Dit is echter een kleine verbetering ten opzichte van de anatomische MRI alleen, en het is niet duidelijk of dit klinisch relevant kan zijn.Show less
Introduction Machine-learning (ML) models trained on MRI data provide promising results for Alzheimer’s Disease (AD) diagnosis. However, extracting features from an MRI scan can take up to 25 hours...Show moreIntroduction Machine-learning (ML) models trained on MRI data provide promising results for Alzheimer’s Disease (AD) diagnosis. However, extracting features from an MRI scan can take up to 25 hours, due to the multitude of feature types, and different software programs used for calculating all these types. To make ML-based AD classification more accessible to doctors and researchers, this study aims to use Feature Reduction (FR) techniques to reduce the number of features needed to accurately classify AD. Therefore, the thesis question is: 'Can feature selection techniques reduce the amount of data needed to accurately classify Alzheimer’s disease from multimodal MRI scans?'. Materials and Methods For this study, 76 clinically diagnosed AD patients and 173 healthy elderly control subjects were scanned. The structural MRI scans were processed using three different programs. Cortical Thickness, Area, and Curvature data were all obtained using the Freesurfer software. Grey Matter Density was obtained using FSL VBM, and subcortical volumes were obtained using FSL FIRST. The functional MRI scans were used to calculate connectivity between resting state networks using a pipeline of different software programs. The data resulting from the four different software programs / pipeline were analysed separately. Three different classification algorithms were used: Elastic Net Regression, Gradient Boost Machines, and Random Forests. The three different algorithms individually went through Recursive Feature Reduction. Here, the models are trained and tested on the data with 5-fold cross validation and get an AUC score, after which the least important feature is removed. This is repeated until there is there is only one feature left. Results The ENR model paired with the data from Freesurfer gave the overall best scores. With 100% of data, the AUC score was only 0.899. With just 5% of data, 10 features, the AUC score improved to 0.952. The maximal AUC score was 0.981, with 18% of data. Thus, it was found that feature reduction not only preserved AUC scores, but improved them significantly. This was true in all cases, with the improvement ranging from 2%-28% when trained on only 5%-30% of features. Furthermore, when training and testing on just 5% of features, the AUC scores had a 6.6% decrease to a 20% increase. Discussion This study has found using Recursive Feature Elimination (RFE) not only preserves AUC scores, but often increases them significantly, ranging from 2% to 28%. This result exceeded the expectations of the hypothesis. Additionally, it was found that training on only 5% of features still preserved and sometimes even increased the height of AUC scores. This means that the goal of the study was exceeded, as feature reduction not only preserves, but improves accuracy scores when classifying AD. Furthermore, it was found that using an Elastic Net Regressor on the 10 outlined features of the Freesurfer Program seems most effective when classifying AD. These findings have external validity, because the areas associated with these features are also commonly linked to atrophy in AD.Show less
Uit onderzoek is gebleken dat de functionele-connectiviteitafwijkingen, aantoonbaar met FMRI-scans, mogelijk eerder zichtbaar zijn dan hersenatrofie en cognitieve achteruitgang bij mensen lijdend...Show moreUit onderzoek is gebleken dat de functionele-connectiviteitafwijkingen, aantoonbaar met FMRI-scans, mogelijk eerder zichtbaar zijn dan hersenatrofie en cognitieve achteruitgang bij mensen lijdend aan de Ziekte van Alzheimer (ZVA). FMRI-scans zouden een toevoeging kunnen zijn op het gebruik van structurele MRI-scans bij het classificeren van de ZVA, specifiek in een vroeg stadium van de ziekte. We hebben dit onderzocht door gebruik te maken van structurele en functionele MRI-scans van 77 mensen met ZVA (MMSE= 20,4 ± 4,5) en 173 cognitief normale ouderen (MMSE= 27,5 ± 1,8). De groep met de ZVA is door middel van een mediaansplit op de MMSE-score gesplitst: participanten met een MMSE-score > 21 zijn geclassificeerd als ZVA-patiënten in een vroeg stadium en participanten met een MMSE-score < 21 zijn geclassificeerd als ZVA-patiënten in een later stadium. De volgende structurele MRI-predicatoren zijn gebruikt: subcorticale volumes, corticale dikte, en grijze-stofdichtheid. De functionele MRI-predicatoren waren: functionele-connectiviteitmatrices en de dynamiek van de functionele-connectiviteitmatrices. De structurele en functionele MRI-predicatoren zijn gescheiden van elkaar en gecombineerd gebruikt in een logistische regressie met een LASSO-penalty. Uiteindelijk werden er verschillende modellen ontwikkeld: een model uitsluitend gebaseerd op structurele MRI-maten, een model gebaseerd op enkel FMRI-maten en een gecombineerd model. Deze modellen zijn vervolgens gebruikt om de waarschijnlijkheid te berekenen dat de deelnemers tot één van de volgende groepen behoorden: ZVA, ZVA in een vroeg stadium, ZVA in een laat stadium, of de controlegroep. Er zijn Receiver Operating Curve plots (ROC) gemaakt en de Area Under the Curve (AUC) zijn berekend om de classificatieprestaties van de modellen te evalueren. Uiteindelijk presteerde het model gebaseerd op structurele MRI-maten het best met een AUC variërend van 0,91 tot 0,98. Het gecombineerde model bestaande uit zowel structurele als functionele MRI-maten behaalde aanzienlijk lagere AUC-waardes variërend van 0,88 tot 0,93. De conclusie van dit onderzoek is dat het toevoegen van functionele MRI-maten aan structurele MRI-maten geen verbetering oplevert in de classificatie van de ZVA in zowel een vroeg als laat stadium van de ziekte.Show less
Het is voor behandeling en medicijnonderzoek van belang om goede biomarkers te vinden voor de ziekte van Alzheimer. Atrofie op structurele MRI scans is zo een biomarker. Functionele connectiviteit ...Show moreHet is voor behandeling en medicijnonderzoek van belang om goede biomarkers te vinden voor de ziekte van Alzheimer. Atrofie op structurele MRI scans is zo een biomarker. Functionele connectiviteit (FC) van het Default Mode Network (DMN) via functionele MRI scans kan echter een vroeger waarneembare biomarker zijn. Deze studie heeft dit onderzocht met als hypothesen dat fMRI beter alzheimer kan classificeren dan sMRI in een vroeg stadium, en dat sMRI beter kan classificeren dan fMRI in een laat stadium van alzheimer. Hiervoor is een classificatieanalyse gedaan met als maten voor de structurele MRI: subcorticale volumes, corticale dikte en grijze stof dichtheid (130 predictoren); voor de functionele MRI: functionele connectiviteit van het DMN met 19 andere resting state netwerken, gedefinieerd aan de hand van een Independent Components Analysis (19 predictoren). De classificatieanalyse is gedaan voor zes modellen, waarvan drie voor vroeg stadium en drie voor laat stadium alzheimer. Voor beide stadia zitten in het eerste model de sMRI maten, in het tweede de fMRI maten en in het derde allebei de MRI maten gecombineerd. De steekproef bevatte 249 participanten, waarvan 173 controles en 76 alzheimerpatiënten. De alzheimerpatiënten zijn opgedeeld in vroeg en laat stadium op basis van alzheimerduur en MMSE score en bestond respectievelijk uit 39 en 37 participanten. Uit de resultaten blijkt dat functionele MRI op zichzelf voor zowel vroeg als laat stadium (AUC resp. = 0.46 en 0.71) slechter classificeert dan structurele MRI (AUC resp. = 0.84 en 0.89). Hiermee is de eerste hypothese verworpen en de tweede bevestigd. Opvallend was wel dat het model met zowel structurele als functionele MRI maten beter classificeerde dan structurele MRI op zichzelf voor vroeg stadium data (AUC resp. = 0.86 en 0.84). Hoewel het een kleine verbetering is, suggereert dit wellicht dat FC iets toe kan voegen aan een gecombineerd classificatiemodel voor vroeg stadium alzheimer.Show less
The ability to correctly classify an individual’s clinical status could help pave the way for early treatment programs for disorders and illnesses of mental- and physical nature alike. Neuroimaging...Show moreThe ability to correctly classify an individual’s clinical status could help pave the way for early treatment programs for disorders and illnesses of mental- and physical nature alike. Neuroimaging data could serve as a basis in reaching this goal of accurate classification. The use of said data, however, does come with challenges. A prominent one of which is the fact that neuroimaging data is highly dimensional, meaning that the amount of features largely exceeds the number of subjects within the data set. Furthermore, research has indicated that the use of heterogeneous, but complimentary, data derived from multiple modalities can be an asset to a model used in a classification setting (Zhang et al., 2010; Schouten et al., 2016). The challenge arises in how to combine the different modalities within a single model. A solution could be the use of machine learning algorithms which search for patterns in the data to draw conclusions. Current literature is, however, lacking meaningful comparisons between different machine learning techniques. Within this project, three different algorithms (support vector machines, Gaussian process classification and multiple kernel learning) have been selected to get insight into (1) whether or not machine learning is able to cope with challenges in the use of neuroimaging data, (2) the difference in performance between these methods and (3) whether or not the use of multiple modalities leads to better results in classification. To this end, 16 alcohol-dependent respondents have been selected along with 32 age-matched healthy controls and have been subjected to both MRI and PET. Models have been trained on data from both separate modalities and on data combining the two modalities. The performances of the models have been assessed by leave-one-subject-out cross-validation and expressed in balanced accuracy and area under the curve. Results indicate that the chosen methods are effective in overcoming challenges arising in the use of neuroimaging data as a means of classification. High balanced accuracies have been found ranging from 76.56% (GPC using PET data) to 100% (GPC using MRI data). Different situations are cause for different solutions and the right choice of algorithm seems to be dependent on, for instance, the fact if either unimodal- or multimodal data is used. Also, settings/optimization of parameters within the model can make a large impact on accuracy. It is therefore advised that researchers try different algorithms and settings before selecting a technique. Different options need to be weighed in order to receive the best possible outcome.Show less
Research master thesis | Psychology (research) (MSc)
closed access
Introduction Frontotemporal dementia (FTD) is a young-onset neurodegenerative disorder with treatments still being in development. For trials testing such treatments, sensitive instruments to...Show moreIntroduction Frontotemporal dementia (FTD) is a young-onset neurodegenerative disorder with treatments still being in development. For trials testing such treatments, sensitive instruments to assess treatment effects are essential. This exploratory study aimed to identify such instruments by investigating gene-specific, presymptomatic cognitive decline and the underlying neural mechanisms of this decline. Methods We examined longitudinal cognitive decline using mixed effects models with natural cubic splines in six different domains for carriers of genetic mutations in GRN (n=46), MAPT (n=22), C9orf72 (n=29), and healthy controls (n=84). A voxel-based morphometry analysis was used to correlate cognitive decline to grey matter volume decline for the three mutation carrier groups. Results MAPT and C9orf72 mutation carriers showed a steeper decline on language (χ2(6) = 21.78, p = .001) and memory (χ2(6) = 18.42, p = .005) compared to GRN mutation carriers and controls. Decline in executive functions was associated with larger grey matter volume decline in the left superior and right middle frontal gyrus for C9orf72 mutation carriers and decline in language was associated with larger grey matter volume decline in the right anterior insula for MAPT mutation carriers. Discussion This study provides evidence of gene-specific cognitive decline in presymptomatic genetic mutation carriers of FTD. The findings highlight the importance of both neuropsychological and neuroimaging assessment which can be used as sensitive diagnostic biomarkers to identify and track disease progression in genetic FTD.Show less
Objective: Multiple sclerosis (MS) is the most common neurodegenerative disease among young adults, of which 40-70% of the patients suffer from cognitive impairment. Currently, there is no...Show moreObjective: Multiple sclerosis (MS) is the most common neurodegenerative disease among young adults, of which 40-70% of the patients suffer from cognitive impairment. Currently, there is no biomarker predicting the cognitive status of MS patients. This study performed a principal component analysis in order to find a disease pattern that can aid in the differentiation of cognitive impairment in MS. Methods: A principal component analysis (PCA) was performed to create a disease pattern based on differences in whole-brain voxel intensities of conventional MRI sequences (T1, T2, and T2- FLAIR) and magnetization transfer (MT)-based MRI of 15 cognitively preserved MS patients (MSCP), 15 impaired patients (MS-CI) and 15 controls. A leave-one-out approach was used to validate the disease patterns between different cognitive performance statuses. Results: None of the conventional MRI sequences nor MT-based MRI were able to find a significant disease pattern for separating MS patients on cognitive status. The frontal cortex, periventricular zone, longitudinal fasciculus, thalamus and brainstem were more severely affected in cognitive impaired MS patients, although significance was not reached. Conclusion: Although the brain patterns created with both conventional MRI sequences and MTbased MRI sequences for evaluating cognitive performance in MS were not significant, the PCA is still a promising technique, when a larger sample size can be included.Show less
In this thesis, the potential of Chemical Exchange Saturation Transfer to image brain metabolites is presented. The Bloch-McConnell equations are simulated and the optimal parameters are found to...Show moreIn this thesis, the potential of Chemical Exchange Saturation Transfer to image brain metabolites is presented. The Bloch-McConnell equations are simulated and the optimal parameters are found to be 3.5uT, tsat = 1s for glutamate and 3uT, tsat = 1.5s for creatine. Furthermore, multiple quantification methods, including MTR asymmetry, Lorentzian fits and spinlock fits are evaluated for quantifying CEST signal from glutamate and creatine. The quantification methods are tested on the Bloch-McConnell simulations, 2-pool phantoms, 3-pool phantoms and in vivo.Show less
In this thesis, we address our progress to send high currents and generating high magnetic fields at milliKelvin temperatures for the use in MRFM measurements. Multiple ways for sending a current...Show moreIn this thesis, we address our progress to send high currents and generating high magnetic fields at milliKelvin temperatures for the use in MRFM measurements. Multiple ways for sending a current while at 20mK inside a dilution refrigerator are described. The use of a heatsink and an option for splitting the current over multifillament connections are analyzed and tested. We find starting resistances in our spotwelds contradicting with earlier measurements of 3pΩ conducted in our group. Next, the design of a transformer in the form of a cone complement is showed and preliminary tests are presented. Furthermore, the inductance is calculated from a sweep over a frequency range from 500Hz to 20kHz. Our measurements show high potential for an experiment to generate 500mT at 20mK. This experiment is described and in addition, a possible use for B1-fields of this cone complement coil is briefly discussed.Show less
Phase-based conductivity mapping using MRI data contains an assumption of locally constant complex permittivity and use of a differential operator which result in significant inaccuracies at tissue...Show morePhase-based conductivity mapping using MRI data contains an assumption of locally constant complex permittivity and use of a differential operator which result in significant inaccuracies at tissue boundaries and amplification of noise in data. This work focuses on the implementation of an iterative model-based nonlinear optimization algorithm that aims to surpass these rising inaccuracies. The algorithm is designed to optimize conductivity maps using phase data acquired from MRI. In addition to optimization, the algorithm focuses on regularization which further improves the optimized outcome of the conductivity maps. Successful results are demonstrated using both simulated as well as phantom data. The comparison between results of a conventional phase-based conductivity mapping and the iterative algorithm shows improved accuracy for the latter. In addition, the model-based algorithm possesses potential for reduced acquisition time as it is capable of reconstructing accurate conductivity maps with relatively low SNR. In the future, experiments on in-vivo data can be performed. Additionally, to improve the accuracy of the conductivity maps even further, implementation of optimal methods to determine regularization parameters and regularization functions is possible.Show less
The Specific Absorption Rate (SAR) is a limiting factor to all MRI-scans. Especially at ultra-high magnetic fields (≥ 7 Tesla), it imposes a significant constraint in the design of pulse sequences....Show moreThe Specific Absorption Rate (SAR) is a limiting factor to all MRI-scans. Especially at ultra-high magnetic fields (≥ 7 Tesla), it imposes a significant constraint in the design of pulse sequences. Due to interpatient variability and the complicated structure of human anatomy, it is difficult to accurately determine the exact SAR-distribution for individual patients. Computational simulations using high-resolution human body models can be used to estimate the SAR, but such models are not available for individual patients in a clinical setting. Here, a method for developing a personalized model for estimating SAR in the head using parallel transmission at 7 Tesla is proposed based on clustered segmentation of tissues. We found that by segmenting all the tissues in the head into fat, cerebrospinal fluid (CSF), grey matter, and bone, the peak-SAR can be determined with an error of less than 2.8 % of the overall peak-SAR. This result is shown to be reproducible for subjects of different ages and genders. Methods for the automated segmentation of this mapping in individual patients based on T1w-images, quantitative T1-mapping, and ultra-short TE-scans are proposed and tested experimentally. Using the proposed method, it should be possible to operate scanners closer to the true SAR-limits due to improved estimations of the actual patient-specific SAR.Show less
In this thesis we investigate conductivity changes due to magnetite in agarose gels mimicking grey brain matter. We use conventional MRI sequences to acquire B + 1 phase maps. Using the homogeneous...Show moreIn this thesis we investigate conductivity changes due to magnetite in agarose gels mimicking grey brain matter. We use conventional MRI sequences to acquire B + 1 phase maps. Using the homogeneous Helmholtz equation and the B + 1 phase-only approximation, we reconstruct conductivity maps. The current sensitivity of the reconstructions is too low to detect conductivity changes due to magnetite nanoparticles in the concentration found in the brain of Alzheimer’s disease patients. Nevertheless, we have promising indications that we have been able to observe a change in the standard deviation of the conductivity due to the presence of magnetite.Show less
In this study, we employed several methods to characterize iron-oxide nanoparticles using SQUID magnetometry and MRI. With SQUID magnetometry, we measured the Isothermal Remanent Magnetization of C...Show moreIn this study, we employed several methods to characterize iron-oxide nanoparticles using SQUID magnetometry and MRI. With SQUID magnetometry, we measured the Isothermal Remanent Magnetization of C. Elegans and two human brain samples. We obtained the iron concentration from the fit. We were able to detect changes in iron concentration due to mutations in C. Elegans. For the MRI measurements, we used Quantitative Susceptibility Mapping and an Off-Resonance Saturation method for brain phantoms. These phantoms consist of different concentrations of magnetite or ferritin dissolved in an agarose gel and mimics the human brain. With QSM we observed a comparable slope of the susceptibility/µg iron/ml. For the ORS method, a good agreement is found between the obtained iron concentration and the pre-determined iron concentration in the sample.Show less
The goal of the project is to assess whether the Off-Resonance Saturation (ORS) method is able to differentiate and quantify different mineralized iron forms, in particular magnetite and ferritin....Show moreThe goal of the project is to assess whether the Off-Resonance Saturation (ORS) method is able to differentiate and quantify different mineralized iron forms, in particular magnetite and ferritin. Samples containing agarose and iron nanoparticles will be prepared and studied with a pre-clinical 7T MRI scanner at the LUMC. First, the samples will be characterized with commonly-used MRI sequences to obtain relaxation time maps and spectra. The main part of the project is to apply the ORS method to acquire positive contrast. Different nanoparticles will be used and the parameters of the ORS sequence will be optimized. Finally, a simulation is made to verify the validity of the ORS theory.Show less
The medical technique of Magnetic Resonance Imaging (MRI) is barely available in developing countries because of its high cost and the strong requirements on infrastructure. To address this problem...Show moreThe medical technique of Magnetic Resonance Imaging (MRI) is barely available in developing countries because of its high cost and the strong requirements on infrastructure. To address this problem, we are developing a permanent magnet-based head scanner that is affordable (<50,000 EUR) and portable. Here, we report on the first observations of magnetic resonance in our custom magnet array with a field strength of 59 mT. Using custom made volume coils, we observe using Hahn echo (Spin echo) and CPMG pulse sequences. We discuss the step towards 2D imaging using rotating spatially encoding magnetic fields (rSEMs) and show simulations that indicate this is feasible in our setup. Finally, we discuss the technical challenges that still have to be overcome to turn this prototype into a diagnostic device for those in need.Show less
The MRI time constant T2 of water is an indication of inflammation in muscles but is difficult to identify when fat is infiltrated in the muscle. A model that takes both the T2 of fat and the T2 of...Show moreThe MRI time constant T2 of water is an indication of inflammation in muscles but is difficult to identify when fat is infiltrated in the muscle. A model that takes both the T2 of fat and the T2 of water into account can be used to determine the T2 of water when the T2 of fat is known. The T2 of fat in affected muscles is equal to subcutaneous fat. To determine the T2 of subcutaneous fat the Extended Phase Graph model is used which calculates the contribution of spins to the MRI signal taking the effects of relaxation, dephasing and radiofrequency pulses into account. It is therefore able to keep track of the contribution of stimulated echoes to the MRI signal. The observed T2 fat values show differences in the order of magnitude of 10 ms in different parts of subcutaneous fat. T2 fat values determined by the presented EPG model are 50 to 80 ms higher than the actual values. Suggested improvements to the EPG model includes usages of a different shaped RF pulse and to take into account different gradient strengths, frequencies of spins and the chemical shift of water and fat.Show less
In developing countries there is a high demand for medical diagnosis and treatment. A diagnosis is generally made with the aid of modern imaging techniques. However these techniques are very...Show moreIn developing countries there is a high demand for medical diagnosis and treatment. A diagnosis is generally made with the aid of modern imaging techniques. However these techniques are very expensive and not available for the majority of people living in third world countries. For the diagnosis of hydrocephalus the requirements of the resolution of an image are not so high. We take the criteria to diagnose young children in developing countries suffering from hydrocephalus as the starting point of the development of a portable low field MRI scanner based on permanent magnets. We also have to take into account that temperature control might be poorly available or even absent in areas where the portable MRI will be used. In this bachelor research project we use measurements of the temporal stability and thermal stability of the Halbach magnet to show what the three-dimensional stability of the magnetic field looks like under temperature fluctuations. This data can be used to create a feedback loop that corrects for the magnetic field drift and eventually results in better image reconstruction.Show less
Low-Field MRI systems without gradients have problems with spatial encoding and low SNR due to the low magnetic field (B0). In this project, an 8 element RF coil array is constructed for extra...Show moreLow-Field MRI systems without gradients have problems with spatial encoding and low SNR due to the low magnetic field (B0). In this project, an 8 element RF coil array is constructed for extra spatial encoding. This thesis describes the idea of Low-Field MRI, how 2D images can be constructed inside its field, the construction of coils and how these coils must be tuned and matched.Show less
B0 magnetic field non-uniformity is the cause of a large amount of image artifacts in MRI. B0 inhomogeneities arise due to magnetic susceptibility differences between tissues. In particular, the 9...Show moreB0 magnetic field non-uniformity is the cause of a large amount of image artifacts in MRI. B0 inhomogeneities arise due to magnetic susceptibility differences between tissues. In particular, the 9 ppm magnetic susceptibility difference between air and tissue generate disturbances in the B0 main field near the skin. We study the B0 passive shimming approach of covering the skin with a susceptibility-matching material from both an experimental and a mathematical viewpoint. In the experimental study, a lightweight and simple to shape pyrolytic graphite composite foam is used to compensate for the field inhomogeneities in the region of the neck. We experimentally demonstrate that the pyrolytic graphite foam improves the uniformity of the static field in a phantom and in vivo at 3T. In the numerical study, we aim for a design of a neck shim which efficiently homogenizes the B0 field while being practically implementable. We propose a level set optimization method as an approach to find the optimum design for a neck shim. Simulations prove that the proposed method is able to solve the topological optimization problem while preserving the imposed constraints.Show less
