Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of Graph Search.
Magnetic resonance imaging (MRI) is highly susceptible to subject's motion and can significantly degrade image quality. In brain MRI exams, involuntary head movements can affect the sampled k-space data. Such unintended alterations may result in visible image artifacts such as blurring, ghosting and others, and therefore potentially disqualify the image from diagnostic purposes. Methods to characterize motion in order to mitigate its impact on image quality exist and range from MR sequence based techniques to scanner independent tracking systems. Although, many motion detection and correction strategies have been proposed in the past, a universal solution to the problem does not exist yet. The work of this thesis was focused on the exploitation of the motion information from a multi-channel Free Induction Decay Navigator (FID) to develop and to optimize motion detection and correction methods in structural brain MRI. After a short introduction to the motion problem in MRI the fundamental methodology behind FID based motion detection is presented and used in this thesis. Considerable work has already been done in the field of motion correction for MRI that is summarized by reviewing the most recent literature, which allowed to reveal some pitfalls in the present approaches and to demonstrate the motivation behind an FID-based method for motion correction in MRI. The first study was conducted to demonstrate that substantial motion information is contained in the multi-channel FID signal, whereby the FID signal is correlated with motion parameters that were obtained from a highly accurate optical tracking system. This work was able to confirm the theoretical foundations from the Biot-Savart law, however, also revealed that a pure FID-based method is not sufficient to exactly calculate the underlying motion trajectory. It is speculated that scanner and subject related information might lead to a closed form solution, yet it was not possible to derive one due to a high dimensionality of the motion problem. Therefore, two alternative approaches were developed to utilize the FID signal for motion detection and correction in MRI. First, a prospective motion correction strategy for an MP-RAGE sequence is demonstrated, whereby the FID signal is used to trigger a prospective motion correction mechanism. The second alternative approach describes how the FID signal can be used to evaluate the quality of an already acquired image and how the FID signal can be used as an optimizer for an autofocusing based retrospective motion correction technique.
Mayeul Sylvain Chipaux, Hoda Shirzad