Chemical Exchange Saturation Transfer MR Fingerprinting

摘要 我们建议开发一种快速、定量的化学交换饱和转移(CEST)成像技术，通过 将CEST与磁共振指纹识别(MRF)和深度学习技术集成在一个统一的框架中， 转化为常规临床实践的最终目标。CEST成像是一种重要的分子磁共振成像方法 这可以基于溶质不稳定质子和散体水质子之间的质子交换来产生对比 在组织里。酰胺质子转移成像(APT)是基于CEST的分子磁共振成像的一种变体，它是基于酰胺的 组织中内源可移动蛋白和多肽的质子(-NH)。APT-MRI已成功应用于 图像蛋白含量和pH，有助于肿瘤分级和活动性复发肿瘤与 治疗效果。然而，目前使用的大多数APT成像协议依赖于定性、 所谓的APT加权(APTw)图像，将检测灵敏度限制在定量参数，如pH 或蛋白质浓度。目前，定量APT成像通常是通过评估所谓的 Z谱，通过测量归一化水信号强度作为饱和频率的函数而产生 在不同的射频(RF)饱和功率下进行补偿，这是非常耗时的。因此，发展 需要快速、定量的APT成像技术。磁流变液是一种新的定量成像方法， 使用伪随机获取参数同时量化多个组织属性，因此， 与传统技术相比，显著提高了扫描效率。MRF已成功应用 在患者研究中评估MR松弛时间T1和T2的范围和变化，提供初始 它的临床效用的证据。深度神经网络的最新进展为有效解决问题提供了新的可能性 MRF重建中的一般反演问题，并在以下位置产生高质量的组织参数估计 速度很快。我们的假设是，通过结合APT、MRF和深度学习技术，我们可以高度 加速图像采集，准确估计组织的量化值。我们的假设将 通过三个具体目标进行测试：1)开发快速3D APT-MRF序列并设计优化的射频 使用深度学习的饱和时间表；2)量化绝对酰胺质子浓度和交换 使用卷积神经网络的比率；以及3)展示该技术在 脑癌，这将通过放射摄影引导的立体定向活检来确认。通过量化APT 成像技术，对胶质瘤的pH和蛋白质含量的先验知识可能有助于对 为患者提供个性化的治疗策略，并帮助监测治疗反应。

ABSTRACT We propose to develop a fast, quantitative chemical exchange saturation transfer (CEST) imaging technique, by integrating CEST with MR fingerprinting (MRF) and deep-learning techniques in a unified framework, with the ultimate goal of translation into routine clinical practice. CEST imaging is an important molecular MRI method that can generate contrast based on the proton exchange between solute labile protons and bulk water protons in tissue. Amide proton transfer (APT) imaging, a variant of CEST-based molecular MRI, is based on the amide protons (-NH) of endogenous mobile proteins and peptides in tissue. APT-MRI has been used successfully to image protein content and pH, enabling tumor grading and the differentiation of active recurrent tumor from treatment effects. However, most currently used APT imaging protocols depend on the acquisition of qualitative, so-called APT-weighted (APTw) images, limiting the detection sensitivity to quantitative parameters, such as pH or protein concentration. Currently, quantitative APT imaging is often attempted by assessing a so-called Z-spectrum, generated by measuring the normalized water signal intensity as a function of saturation frequency offset under varied radiofrequency (RF) saturation powers, which is time-consuming. Thus, the development of fast, quantitative APT imaging techniques is needed. MRF is a novel quantitative imaging method that simultaneously quantifies multiple tissue properties using pseudorandom acquisition parameters, and thus, significantly improves scan efficiency compared to conventional techniques. MRF has been successfully applied in patient studies to evaluate the range of and changes in MR relaxation times, T1 and T2, providing initial evidence of its clinical utility. Recent advances in deep neural networks open a new possibility to efficiently solve general inversion problems in MRF reconstruction, and to produce high-quality estimates of tissue parameters at high speed. Our hypothesis is that, by combining APT, MRF, and deep-learning techniques, we can highly accelerate image acquisition and accurately estimate the quantitative values of the tissue. Our hypotheses will be tested through three specific aims: 1) to develop a fast 3D APT-MRF sequence and design an optimal RF saturation schedule using deep-learning; 2) to quantify absolute amide proton concentrations and exchange rates using convolutional neural networks; and 3) to demonstrate the initial clinical utility of the technology in brain cancer, which will be confirmed by radiographically-guided stereotactic biopsy. Through quantitative APT imaging technology, a priori knowledge of the pH and protein content in gliomas may help in the stratification of patients into personalized therapeutic strategies and help monitor treatment response.