Optimisation of rapid acquisition techniques in 1H MRSI for the potential to obtain complex metabolite data at higher resolution in a clinically feasible time

Project completed 2014.

Supervisors:
Dr Nigel Davies, UHB
Dr Andrew Peet, Institute of Cancer and Genomic Sciences
Dr Martin Wilson, BUIC

Brain tumours are the most frequent form of solid cancer in children, with around 400 cases per year in the UK. Compared with those in adults, childhood brain tumours exhibit a greater diversity of types and grades of malignancy, with a larger range of clinical outcomes.The majority of childhood brain tumours are classified as low grade with relatively good survival rates, however the effects of both the tumour and treatment often result in significant long-term disability. A high proportion of childhood brain tumours occur in parts of the brain where surgery or even biopsy is high risk, such that histopathological diagnosis is not possible.This leads to significant challenges in the clinical management of these tumours and places greater importance on non-invasive diagnosis.An example of this is a group of tumours known as Optic Pathway Gliomas (OPG).This is a type of low grade glioma, which arise from the glial cells that form a supportive network for the brain's neurons. Due to proximity to the optic nerve pathway, OPG often leads to visual loss or blindness.

Magnetic resonance imaging (MRI) is a non-invasive technique used in the routine clinical assessment of brain tumours in general and paediatric brain tumours in particular.While conventional MRI provides excellent delineation of structural and gross pathological changes, it has limited sensitivity to pathology at a cellular level, and tissue infiltration beyond the tumour mass goes undetected. Magnetic Resonance Spectroscopic Imaging (MRSI) is a functional MRI technique allowing the non-invasive imaging of cellular metabolism and improved characterisation of brain tumour heterogeneity. Currently, clinical applications of this technique are limited by its low resolution (~1cm3) and long acquisition times. Faster or higher resolution MRSI would be of great benefit for a range of clinical applications, including the assessment of OPG.

Achieving enhanced resolution MRSI requires a method of rapid acquisition whilst maintaining adequate point spread function (PSF), signal-to-noise ratio (SNR) and minimizing aliasing and lipid contamination. The result of poor PSF in an MRSI context is often referred to as 'bleed-through'.The aim of this project is to develop an optimized rapid acquisition pulse sequence for MRSI at Birmingham Children's Hospital that permits resolution enhancement within a clinically feasible time frame. Compressed Sensing (CS) is a promising method of reconstructing sparse datasets without unduly compromising on data quality. CS has recently been successfully used to speed up conventional MRI, but research into its application in MRSI (CS-MRSI) is limited and its impact on spectral image quality is relatively unexplored.

The prime objective of this project is to investigate the effect of CS on spectral image quality in 2D phase-encoded MRSI. CS variables including type of sparse transform used, variable density sampling strategy and acceleration factor have been optimised by measuring their effect on spatial resolution, SNR and lipid contamination in MRSI.Software for reconstruction of retrospective MRSI data using CS and optimization of the technique has been developed. Comparison of CS-MRSI with results from fully-sampled conventional reconstruction in phantoms, simulations and in-vivo data has been used to determine quality measures. Construction and scanning of a phantom containing a sharp metabolite boundary allowed direct assessment of bleed-through and measurement of the edge-response function.A software phantom was developed to investigate higher resolution MRSI datasets in a highly controllable manner. This phantom allowed the effect of SNR, nominal voxel size and spatial varying metabolite data to on the results of retrospective reconstruction with compressed sensing to be investigated.The research is to be concluded by application of CS-MRSI on retrospective OPG patient data to define its advantage in a real, clinical scenario.

Link to ethesis: http://etheses.bham.ac.uk/6128/