WAVELET NOISE REDUCTION AND VASCULAR WATER TRANSPORT MODELLING : APPLICATIONS TO DIFFUSION AND PERFUSION MRI

Abstract: Magnetic resonance imaging (MRI) is a powerful medical imaging technique, used to detect and characterise a range of diseases and conditions. It is based on the use of a strong static magnetic field in combination with magnetic field gradients and pulsed radiofrequency electromagnetic fields to visualise various organs and structures in the body according to their morphology or function.Diffusion and perfusion MRI are established methods for quantitative measurements, often used in neurological and neurovascular clinical applications. Although these techniques are often used separately to investigate a number of diseases, combined diffusion and perfusion information can provide unique information, e.g. for assessment of whether stroke patients in the acute stage are likely to benefit from reperfusion therapy. This may be accomplished by identification of the so-called ischemic penumbra (i.e. the area surrounding the core of an infarct, exhibiting disturbed microcirculation but still viable and salvageable if the local blood supply is efficiently restored). This identification concept is often referred to as the diffusion–perfusion mismatch. In oncological applications, a combination of diffusion and perfusion MRI is sometimes used in tumour characterisation and in attempts to monitor early treatment response.Quantitative diffusion MRI may be hampered by a bias induced by the so-called rectified noise floor in areas with low signal-to-noise ratio (SNR), and to address this issue, a wavelet-based filtering method was presented and used for noise reduction in diffusion MRI.Perfusion images acquired by arterial spin labelling (ASL), which is the technique investigated in the present work, suffer from inherently low SNR, and this is commonly addressed by averaging multiple repetitions, which leads to a prolonged acquisition time. As an alternative approach, wavelet-domain filtering for noise reduction was applied to ASL data, and the performance of the proposed filtering technique was investigated (in terms of accuracy, precision and structural degradation), and a comparison with conventional Gaussian smoothing was also included. Additionally, a quantitative non-compartment modelling approach for assessment of blood water transit time through the microvasculature and the blood–brain barrier (BBB) was investigated. In one study, the model was adapted to a clinical setup and applied to test–retest data from healthy volunteers, and the effects of noise on the model were examined by simulations. In an animal study, the model was further developed by introducing a bolus- tracking ASL solution that included a measured arterial input function (AIF) instead of a theoretical rectangular input function. Furthermore, it was explored whether effects of mildly damaged red blood cells on microvascular parameters were detectable using the proposed modelling approach and by ASL-based CBF quantification.The extracted water transit time parameters can be used separately or in combination with conventional perfusion and diffusion estimates. Changes in the blood water transit time in the microvasculature may be related to alterations in capillary water permeability, and may thus be useful in the assessment of BBB integrity. Disturbed BBB permeability has been attributed to a number of disease states, and may be relevant to, for example, early diagnosis of Alzheimer's disease, inflammation, tumour grading and ischaemic stroke.