Search

Abstract: Quantitative Susceptibility Mapping (QSM) is an emerging postprocessing method, computed from phase images, which is finding wide application in quantifying iron content in healthy and pathological tissue. However, QSM is still not commonly used in clinical practice. This thesis discusses the challenges that come during the application of QSM to patient studies and makes advances to solve problems such as long acquisition times, motion, and works towards finding more applications for QSM. The focus for this work is on stroke applications where methods such as Susceptibility Weighted Imaging (SWI) and Time-of-Flight (TOF) MR Angiography (MRA) are already used.SWI finds one application in the study of hemorrhage, and it has been shown in previous studies that QSM can be reconstructed from the single echo SWI sequence. However, whole brain SWI requires an acquisition time of about 5 mins which is often too long for hemorrhagic patients to remain still inside the scanner. In Chapter 2, a rapid single-shot Echo-planar Imaging (EPI) sequence with acquisition time of 0.45 mins was applied to subjects with intracerebral hemorrhage (ICH) which enabled rapid measurement of ICH area and mean magnetic susceptibility, with reduced motion as compared to standard SWI. EPI requires minimal additional acquisition time and hence can be incorporated into iron tracking studies in ICH.Motion effects cause artifacts in magnitude as well as phase images. Hence, Chapter 3 investigates the quantitative effects of movement and respiratory fluctuations on QSM in the brain. QSM was found to be more sensitive to motion caused by movement than magnitude images and thus post-processing motion correction or faster sequences may be beneficial for QSM applications. However, respiratory fluctuations did not cause statistically significant differences in susceptibility values in group study; although, these variations might be considered important in individual follow-up studies.SWI is widely used in the study of veins, hematoma, lesions etc. However, since it uses filtered phase for its computation, SWI has certain limitations such as artifacts arising from phase wraps, blooming effects, dependence of phase value on the orientation of object with main magnetic field etc. In order to overcome SWI limitations, a new method called quantitative susceptibility weighted imaging also known as true susceptibility weighted imaging (tSWI), has been recently introduced which uses susceptibility maps instead of filtered phase. Chapter 4 aims at optimizing tSWI parameters for strong susceptibility sources like hemorrhage and investigates the benefits and limitations of tSWI for hemorrhages. In hemorrhage, tSWI minimizes both blooming effects and phase wrap artifacts observed in SWI. However, unlike SWI, tSWI requires an alteration in the threshold limits for best hemorrhage depiction that greatly differs from the standard values. tSWI can be used as a complementary technique for visualizing hemorrhages along with SWI.It is always desirable to obtain maximum information from a single acquisition. Hence in Chapter 5, a new sequence has been introduced to simultaneously compute TOF-MRA, QSM, SWI and transverse relaxation rate R2* while maintaining all the key features of standard TOF-MRA such as multiple overlapping thin slab acquisition (MOTSA), ramped RF pulses and venous saturation. The effect of these TOF features on QSM and SWI was studied. The proposed sequence with the TOF features provided TOF-MRA and SWI with similar CNRs to standard methods. The mean susceptibility values for brain structures had no significant susceptibility variation between the proposed and standard methods as well. Thus, this sequence is able to provide similar TOF-MRA to standard TOF methods while enabling additions of SWI, R2* and QSM.

Abstract: Quantitative susceptibility mapping (QSM) is an emerging magnetic resonance imaging (MRI) post-processing technique that uses phase images from a gradient echo sequence to create magnetic susceptibility maps. QSM is sensitive to paramagnetic iron found in various states of blood, from deoxyhemoglobin to hemosiderin. QSM can also differentiate between calcium and iron deposits. Unlike phase images, QSM is not strongly dependent on object shape nor orientation with respect to the main magnetic field. A 3D flow-compensated gradient echo was used in this work for reconstruction of QSM from phase images. An introduction to MRI, phase images, susceptibility and susceptibility-weighted imaging (SWI) is given in Chapter 1. This chapter will also explain the details of QSM reconstruction and an overview to various applications.Tissue and draining vein oxygenation level in the brain is a prominent parameter in determining the severity of ischemia. Due to its sensitivity to deoxyhemoglobin, QSM provides a quantitative measure of brain tissue and venous oxygenation. QSM can also distinguish between cerebral microbleeds (CMBs) and calcium deposits and provides valuable information of the size and quantity of CMBs. Chapter 2 introduces stroke, with a focus on ischemic stroke and cerebral microbleeds, as the major clinical applications of this research.Recent studies in ischemic stroke using QSM have only examined severe ischemic strokes. Furthermore, no studies to date have examined all three potential applications of QSM to ischemic stroke: cerebral microbleed quantification, and vein and tissue oxygenation. In Chapter 3, the value of QSM in ischemic stroke patients was investigated in two ways. First, by evaluating susceptibility in tissue and draining veins from infarcted and contralateral regions in acute stroke with a wide range of stroke severity; and second, assessing microbleed burden on QSM in comparison to standard SWI in chronic stroke. QSM analysis in acute ischemic stroke revealed significantly greater susceptibility (p< 0.001) for vein measurements in the infarcted hemisphere compared to the contralateral homologous hemisphere. This can be an indicator of higher blood deoxygenation in veins draining from the infarcted hemisphere, suggesting a compromised arterial blood flow and collateral vascular distribution in the ischemic region. Tissue analysis showed a significantly higher susceptibility (p< 0.036) of white matter in infarcted regions compared to healthy regions which can be an indicator of higher deoxygenation in the infarct, while no significant higher susceptibility was observed in grey matter between the two hemispheres. This may be due to higher sensitivity of white matter to ischemia.Microbleed assessment using QSM revealed key advantages of QSM over SWI, such as smaller size of microbleed (p< 0.0001) as well as potentially improved detectability of microbleeds. This provides quantitative measurement of susceptibility rather than the qualitative nature of SWI and removal of blooming signal artifacts originating from strong dipole field effects seen on SWI.In summary, this thesis has demonstrated the value of QSM in ischemic stroke for measuring both oxygenation and microbleeds. Since QSM can be obtained from raw phase images of standard SWI sequences, no extra scan time is needed to enable QSM studies when SWI is already performed.

Magnetic resonance imaging (MRI) is a high-resolution, non-invasive medical imaging modality that is widely used in human brain. In recent years, susceptibility weighted imaging (SWI) and quantitative susceptibility mapping (QSM) have been proposed to utilize MR phase signal to generate contrast from tissue magnetic susceptibility and even quantify the property. On the other hand, deep learning, especially deep convolutional neural networks (DCNNs), have achieved state-of-the-art performances in numerous computer vision tasks and gained significant attention in the field of medical imaging in the recent years. This dissertation combined the idea of deep learning with the two MR phase imaging methods. To combined deep learning with SWI, we designed and trained a 3D deep residual network that can distinguish false positive detected candidates from cerebral microbleeds (CMBs) and built an automatic CMB detection pipeline with high performance. We further confirmed the generalizability of this deep learning-based pipeline using multiple dataset with different scan parameters and pathologies and provided lessons for application and generalization of generic deep learning based medical imaging methods.To combine deep learning with QSM, we developed a 3D U-Net based network that learns to perform dipole inversion from gold standard QSM acquired from data with multiple orientation. The model was further improved with adversarial training strategy and achieved significantly lower reconstruction error than traditional QSM algorithms. In addition, we also performed various background removal and dipole inversion algorithms on both brain tumor patients and healthy volunteers to study and compare their performances. The results could provide guidance on future application of QSM in different scenarios.

Abstract: Magnetic resonance imaging is an essential tool for assessing soft tissues. The desire for increased signal-to-noise and improved tissue contrast has spurred development of imaging systems operating at magnetic fields exceeding 3.0 Tesla (T). Unfortunately, traditional imaging methods are of limited utility on these systems. Novel imaging methods are required to exploit the potential of high field systems and enable innovative clinical studies. This thesis presents methodological advances for human brain imaging at 4.7 T. These methods are applied to assess sub-cortical gray matter in multiple sclerosis (MS) patients. Safety concerns regarding energy deposition in the patient precludes the use of traditional fast spin echo (FSE) imaging at 4.7 T. Reduced and variable refocusing angles were employed to effectively moderate this energy deposition while maintaining high signal levels; an assortment of time-efficient FSE images are presented. Contrast changes were observed at low angles, but images maintained a clinically useful appearance. Heterogeneous transmit fields hinder the measurement of transverse relaxation times. A post-processing technique was developed to model the salient signal behaviour and enable accurate transverse relaxometry. This method is robust to transmit variations observed at 4.7 T and improves multislice imaging efficiency. Gradient echo sequences can exploit the magnetic susceptibility difference between tissues to enhance contrast, but are corrupted near air/tissue interfaces. A correction method was developed and employed to reliably produce a multitude of quantitative and qualitative image sets. Using these techniques, transverse relaxation times and susceptibility field shifts were measured in sub-cortical nuclei of relapsing-remitting MS patients. Abnormalities in the globus pallidus and pulvinar nucleus were observed in all quantitative methods; most other regions differed on one or more measures. Correlations with disease duration were not observed, reaffirming that the disease process commences prior to symptom onset. The methods presented in this thesis enable efficient qualitative and quantitative imaging at high field strength. Unique challenges, notably patient safety and field variability, were overcome via sequence implementation and data processing. These techniques enable visualization and measurement of unique contrast mechanisms, which reveal insight into neurodegenerative diseases, including widespread sub-cortical gray matter damage in MS.