Investigation of a pulsation artefact reduction technique in diffusion-weighted magnetic resonance imaging of the liver

肝脏磁共振弥散加权成像（DWI）的一个问题是左肝叶的信号丢失，这是由心脏搏动引起的。本文的目的是通过减少该伪影来提高肝脏DWI的图像质量，同时实现黑血图像。为此，我们提出了一个混合协议应使用，其可行性在最近的一项志愿者研究中得到了证明。它包含一个传统的双极扩散编码在一个小的b值（50秒/毫米²），和一个流动补偿扩散编码在一个高的b值（800秒/毫米²）。该方案适用于一项前瞻性患者研究，包括已知或疑似肝脏病变的患者。左肝叶病变的可见性是评估的关键参数之一。为了进一步提高图像质量，需要开发和评估先进的后处理方案。它们包括传统的方法，例如异常值检测。此外，需要训练卷积神经网络，使其从获取的扩散加权图像集生成高质量的图像。

One problem in magnetic resonance diffusion-weighted imaging (DWI) of the liver is the signal loss in the left liver lobe, which is caused by cardiac pulsation. The aim of this proposal is to improve the image quality of liver DWI by reducing this artefact and achieving black-blood images at the same time. For this purpose, a mixed protocol shall be used, whose feasibility for this task was proven in a recent own volunteer study. It contains a conventional bipolar diffusion encoding at a small b-value (50 s/mm²), and a flow-compensated diffusion encoding at a high b-value (800 s/mm²). This protocol shall be used in a prospective patient study including patients with known or suspected liver lesions. The visibility of the lesions in the left liver lobe is one of the key parameters that shall be evaluated. To achieve a further improvement of the image quality, advanced postprocessing schemes shall be developed and evaluated. They comprise conventional approaches featuring e.g. outlier detection. Moreover, convolutional neural networks shall be trained to generate high quality images from the set of the acquired diffusion-weighted images.