Show simple item record

dc.contributor.authorWickens, Christopher
dc.date.accessioned2022-04-21T16:07:08Z
dc.date.available2022-04-21T16:07:08Z
dc.date.submitted2021-12-14
dc.identifier.urihttps://www.repository.cam.ac.uk/handle/1810/336329
dc.description.abstractTraumatic brain injury (TBI) is a major health problem worldwide and is the leading cause of mortality and disability of young adults in the developed world. In vivo magnetic resonance spectroscopy (MRS) has proved to be a valuable tool in better understanding the heterogeneous changes in neurochemistry following injury. This thesis is concerned with the development of in vivo MRS methods to support further investigations into TBI. Dynamic 13C MRS is unique in its ability to directly measure in vivo rates of cerebral metabolism non-invasively. In this thesis the 13C-labelled substrate [2-13C]glucose was used for infusion due to its reduced technical demands. To date [2-13C]glucose infusion research has mainly focused on acquisition based methodological development, specifically when observing the carbonyl region. This thesis describes the infrastructural development required to run a robust dynamic 13C MRS infusion study with focus on data acquisition, sequence development, signal processing and mathematical modelling of in vivo metabolic fluxes within the brain. The first attempt at using [2-13C]glucose to quantify the in vivo neuronal tricarboxylic acid (TCA) cycle flux (VtcaN ) in humans is shown, accompanied with a Monte-Carlo error analysis to give a feel for the precision of the derived metabolic fluxes. The measured neuronal TCA cycle fluxes were in line with literature values obtained using different substrates (e.g. [1-13C]glucose). A novel MRS sequence was also built, and tested in vitro, to work at low RF powers while maximising the information content of the experiment. Given the long scan times and extra equipment required in the MRI scanner room for TBI patients the likelihood of artifacts appearing is increased. Two statistical measures to retrospectively measure the magnitude and arrival time of artifacts in the infusion time course are proposed. An understanding of the time at which an artifact arises could be invaluable in trying to understand the root cause of the artifact. Since TBI is a very heterogeneous disorder a wide range of metabolites have been shown to change following injury. Two-dimensional proton (1H) MRS techniques are advantageous in reducing signal overlap and being able to simultaneously acquire both long and short relaxation time metabolites. Specifically, this thesis demonstrates the first in vivo application of adiabatically localised Correlation Spectroscopy (AL-COSY) at 7T to aid in future TBI studies. A script based 2D MRS processing toolbox was derived from the open source 1D FID-A library to support more research in this area. Finally, use of a machine learning approach to provide a direct route to classification of patient outcome following TBI was implemented using previously acquired 31P MRS data. It was found that the machine learning algorithms performed well above chance level. Given the heterogeneous nature of pathology underlying TBI a key feature of machine learning methods is their ability to learn from multiple factors that impact on a patients ability to recover. Medical MRS brain injury data sets are often small, particularly when patients are in a neuro critical condition. Large error bounds are intrinsic to classification problems with small sample sizes and are often overlooked. To correctly account for this short coming a probabilistic view to interpret patient outcome classification results of TBI is promoted. In summary, this thesis explores the technical setup, acquisition and analysis required to perform these technically demanding dynamic 13C MRS infusion studies for use in TBI patients, accompanied with novel methodological contributions to the 13C MRS field. This is followed by the programming of a 2D MRS AL-COSY sequence and its subsequent first in vivo implementation at 7T for future brain injury studies. Finally, I have also presented the first 31P MRS machine learning classification approach to successfully predict between three subject groups: TBI-unfavourable outcome, TBI-favourable outcome and healthy control.
dc.rightsAll Rights Reserved
dc.rights.urihttps://www.rioxx.net/licenses/all-rights-reserved/
dc.subjectTraumatic brain injury
dc.subjectTBI
dc.subjectCarbon-13
dc.subjectinfusion
dc.subjectCOSY
dc.subjectadiabatic
dc.subjectin vivo MRS
dc.subjectMagnetic resonance spectroscopy
dc.subjectPhosphorus-31
dc.subjectMetabolic modelling
dc.subjectMachine learning
dc.subjectclassification
dc.subjectPatient outcome prediction
dc.subjectDynamic carbon-13 infusion
dc.titleMethod Development of 13C, 1H and 31P Magnetic Resonance Spectroscopy for use in Traumatic Brain Injury
dc.typeThesis
dc.type.qualificationlevelDoctoral
dc.type.qualificationnameDoctor of Philosophy (PhD)
dc.publisher.institutionUniversity of Cambridge
dc.date.updated2022-04-08T11:20:43Z
dc.identifier.doi10.17863/CAM.83748
rioxxterms.licenseref.urihttps://www.rioxx.net/licenses/all-rights-reserved/
rioxxterms.typeThesis
dc.publisher.collegeHughes Hall
cam.supervisorCarpenter, Thomas Adrian
cam.depositDate2022-04-08
pubs.licence-identifierapollo-deposit-licence-2-1
pubs.licence-display-nameApollo Repository Deposit Licence Agreement


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record