Assessing the value of a priori information in solving the MEG inverse problem

dc.contributorAalto-yliopistofi
dc.contributorAalto Universityen
dc.contributor.advisorMäkinen, Antti
dc.contributor.advisorParkkonen, Lauri
dc.contributor.authorVranou, Gavriela
dc.contributor.schoolPerustieteiden korkeakoulufi
dc.contributor.supervisorParkkonen, Lauri
dc.date.accessioned2019-06-26T11:08:25Z
dc.date.available2019-06-26T11:08:25Z
dc.date.issued2019-06-17
dc.description.abstractMagnetoencephalography (MEG) and magnetic resonance imaging (MRI) techniques have been steadily advancing, with new methodologies and technical approaches being developed. The challenge posed by MEG is to determine the location of electric activity within the brain from the induced magnetic fields outside the head. In a typical MEG system there are 306 channels, and model there can be thousands of possible source locations, making the corresponding linear system of equations underdetermined. Consequently, the electromagnetic inverse problem remains ill-posed. In this study, a software framework was implemented to simulate data from Neural Current Imaging (NCI), an MRI technique, and the improvement in reliability and accuracy of the MEG inverse solution was investigated when prior information provided by NCI data become reliable. The aim of the thesis was to assess the optimal use of information obtained by NCI, and how the inverse solution accurately guided by NCI improves compared to the state of the art. Combining data from different modalities is expected to provide more information about the underlying process than a single modality, motivating the combination of NCI and MEG. Since both signals are generated by cortical primary currents, a full Bayesian approach was adopted, where NCI is encoded as prior information in the MEG source estimation. Existing open-source software was used for incorporation of NCI information, for the effect of using NCI information to be readily comparable. NCI prior encoding was observed to successfully decrease the influence of activation from other locations and the activity was contained in the relevant cortical locations. However, this effect is directly tied to the image resolution and signal to noise ratio. Some of the degrees of freedom of the simulation framework can be revisited and adjusted once real data become available, in order to provide an accurate forward model for NCI. Various source patterns can be studied to explore the situations in which NCI could provide relevant spatial information for solving the MEG inverse problem. Finally, this framework can be expanded to different source models, volume conductor models and various source localization approaches that could possibly overcome current limitations.en
dc.format.extent45+1
dc.identifier.urihttps://aaltodoc.aalto.fi/handle/123456789/39164
dc.identifier.urnURN:NBN:fi:aalto-201906264229
dc.language.isoenen
dc.programmeMaster’s Programme in Computer, Communication and Information Sciencesfi
dc.programme.majorMachine Learning, Data Science and Artificial Intelligencefi
dc.programme.mcodeSCI3044fi
dc.subject.keywordULF-MRIen
dc.subject.keywordMEGen
dc.subject.keywordNCIen
dc.subject.keywordneuronalen
dc.subject.keywordBayesianen
dc.subject.keywordfusionen
dc.titleAssessing the value of a priori information in solving the MEG inverse problemen
dc.typeG2 Pro gradu, diplomityöfi
dc.type.ontasotMaster's thesisen
dc.type.ontasotDiplomityöfi
local.aalto.electroniconlyyes
local.aalto.openaccessno
Files