Study On Real-Time Motion Correction Method For Magnetic Resoannce Imaging And The Related Methods For Diffusion Imaging

Magnetic Resonance Imaging (MRI) has become an important technique in medical imaging, and it has been widely used in many fields including medicine and cognitive neuroscience research, because it is non-invasive, safe, and provides high tissue contrast which are distinctive and are not available in other imaging modalities for living tissues. While MRI is a relatively young technique, many aspects of it need to be and can be improved. Meanwhile, clinical application of MRI is also fast evolving, and investigators are developing various novel approaches to solve practical problems that they are facing in daily clinical studies. However, the excessive speed of fast development of these solutions naturally brings about a series of technical difficulties that cannot be solved properly in time This dissertation aims to make some contributions along these two axes.In Part1, we study the topic of real-time motion correction using an optical tracking system. This part first introduces the principle of MRI, describes how motion affects imaging quality, and reviews the methods of motion correction that are currently available. We then study in detail our proposed method for motion correction using an optical tracking. We have developed a tracking system for motion in3D space based on a two-camera system, and have applied it to data acquisition on a Siemens3T MR platform, which has been designed for acquiring MR imaging data that are free of motion from moving objects using a technique called "real time correction of head motion". The study shows that solving the motion problem using the optical system is feasible, and this method has the potential to completely avoid motion artifacts in MRI.In Part2, our study focuses on the applications of MRI. In particular, we study a novel diffusion imaging technique, called "Diffusion Kurtosis Imaging (DKI)". Diffusion imaging is an important clinical application of MRI, and it can measure the motion property of water molecules in biological tissue. DKI is a diffusion model proposed in recent years, which can provide new insight on structural information of biological tissues as to the compartment complexity in living tissue. This part first introduces the principle and various models of diffusion imaging, and describes the technical details of the DKI model. Subsequently, we study the following issues pertaining to the DKI model:1) Fast schemas for acquisition of DKI data. This section first reviews various acquisition schemas adopted in previous studies, and qualitatively and quantitatively evaluates a great number of acquisition schemas (varying the involved numbers of b-value and gradient direction) using simulated datasets. We search for schemas of high performance and consequently make recommendations of optimized acquisition schemas. We subsequently test the performance of the optimized schema based on DKI datasets from patients of a stroke study. The conventional analysis methods, such as voxel-based analysis (VBA) and region-of-interest (ROI) analysis, are applied to examine the performance of the recommended schemas. The study shows that the optimized schema yields very similar results in VBA and ROI analysis, compared to those from the conventional schemas. However, the optimized schema reduces the acquisition time by more than a half. It therefore leads to a conclusion that the fast acquisition schema can significantly accelerate the data acquisition process of DKI, which is a desired feature for clinical applications in practice.2) A Fast algorithm for estimation of DKI parameters. In this section, we propose an alternatively iterative method, which adopts an iterative framework to calculate the two parameters of the DKI model, namely apparent diffusion coefficient (ADC) and apparent kurtosis coefficient (AKC), alternatively one after the other. The proposed method converges extremely quickly with great accuracy and precision. It reduces the computational time for a typical DKI dataset from tens of minutes or even several hours to merely one or two minutes, which is a feature particularly suitable for clinical applications. In addition, the algorithm incorporates biological and physical constraints that are popularly agreed, and it also incorporates a smoothing procedure into the iterative framework, which effectively improves the estimation accuracy of DKI parameters, and enhances the visibility of delicate structure in AKC and ADC maps.3) Automatic processing of diffusion data. This section first introduces conventional processing pipeline of diffusion data, and then creates Linux and Matlab scripts to automatically run the processing pipeline. It allows to prescribe all the processing procedures and related parameters in one script program, and to run the program without extra human intervention. Using such a script for data processing facilitates the processing of large amount of imaging data, because the script typically logs all the actions taken on the data, and therefore is easy for modification and future maintenance of the data.In Part3, combining our work in Part1and2, we study the application of our motion correction method to DKI data acquisition. This part attempts to incorporate the optical system for motion correction that we have developed in Part1into the data acquisition process of DKI. The study verifies that in principle the optical system can accurately correct the errors induced by motion in diffusion images and gradient directions of diffusion.