Sparse sampling and constrained reconstruction for accelerated contrast enhanced magnetic resonance imaging

点击翻译按钮获取中文摘要

Magnetic resonance imaging (MRI) is a powerful modality but remains limited. Time-resolved scans, which aim to capture changing physiology, are severely hindered by long scan times. This is particularly true for dynamic contrast enhanced (DCE) MRI, which aims to visualize the flow of an injected contrast agent as it passes through the vasculature. This technique has significant clinical value yet is rarely performed due solely to acquisition limitations. Hardware advances continue to reduce scan times, but with diminishing returns. New methods for accelerated imaging are urgently needed. My research program established by this Discovery Grant will develop, validate, and translate novel methods for highly accelerated dynamic MRI.My research program will explore sparse sampling and constrained reconstruction—together termed compressed sensing (CS)—for faster MRI. These techniques exploit what is well known from the JPEG file format: images can be stored using a fraction of their original data. CS posits that since images can effectively stored, there is no need to acquire all of the original data. This enables image formation from fewer samples than are typically required—thus faster imaging. Compression becomes more effective as dimensions are added, making 4D imaging (i.e., dynamic 3D) a strong candidate for acceleration. My preliminary work has demonstrated that DCE MRI is amenable to extreme speed-ups: acceleration an order of magnitude beyond conventional methods is feasible. Despite its potential, optimal methods for CS are poorly established. Furthermore, the utility of high-resolution dynamic scans has not been explored: the same source data will allow extraction of multiple pieces of physiological information. Currently, no MRI vendor has productized CS—this is due to technological barriers, not utility. It is clear that CS will play a vital role in medical imaging but effective techniques currently preclude widespread adoption.The projects detailed in this application will form the basis of a trainee focused research program with a long-term goal to develop, implement, and commercialize advanced data sampling and reconstruction technologies for MRI. Specific aims of this proposal are:Aim 1: design novel acquisition and reconstruction algorithms for accelerated 4D imaging.Aim 2: develop a comprehensive DCE exam.Aim 3: implement on-line reconstruction of these techniques.The first aim of this project will focus on optimizing data acquisition methods that are amenable to both conventional image reconstruction (albeit with low temporal resolution) and to constrained reconstruction (for high fidelity images). The second aim of this project will exploit the vastly improved image quality available from Aim 1 to extract multiple pieces of clinical data. This is not possible from current DCE scans and has the potential to revolutionize how contrast enhanced imaging is performed clinically. The third aim will strive to tightly integrate the techniques defined by Aims 1 and 2 with the MRI scanner to enable rapid image reconstruction. This objective is often overlooked yet is crucial to facilitate clinical validation and to expedite commercialization.In summary, this application proposes technological innovations to accelerate MRI. Applications of the proposed inventions reach far beyond our initial clinical application (DCE). Trainee involvement, industrial and clinical collaborations, and commercialization will be central themes in my research program.