Development of Methods for a Simplified and Reliable Prostate Cancer MRI Exam

摘要 应用磁共振成像对前列腺癌(PC)的非侵袭性、多参数特征进行研究 (MRI)方法是一个活跃的研究领域，在提供更好的诊断和 治疗监测。由国际团队建立的PI-RADSv2评估体系 的专家承认定量图像在PC诊断中的价值，但在很大程度上依赖于定性 加权图像的评估。虽然这种分级方法达到了相当好的分离效果，但- 在正常和异常的前列腺组织之间，它不能实现适当的惰性分离 和侵袭性疾病，有可能进行更多不必要和昂贵的手术- 给病人的生活质量带来可怕的后果。高价值协议，不需要侵入性和 昂贵的直肠内射频线圈正在研究中。这是以扩展扫描为代价的 时间和在空间分辨率、信噪比和信号偏差方面降低的图像质量，这 对多参数磁共振成像的敏感度和特异度fi产生负面影响。随着这一数字的显著增加 多参数磁共振检查，也有整合的愿望，由最新的革命支持 在诊断成像中，即机器学习中。越来越清楚的是，为了避免不得不 为每个特定的fic系统和协议训练神经网络，可重复性的，因此最好是定量的 成像协议至关重要。为了克服这些限制，我们提出了两种脉冲序列开发-- ADC的有效性和重复性的研究以及新的后处理策略。整体而言 目的是用定量的价值来证明病变特征的附加值，并在 同时了解并最大限度地减少协议选择和扫描硬件的fl影响，从而提高 总体重现性。SPECIfic Aim 1将专注于低失真磁共振成像的开发 用于T2和扩散信号衰减的快速并发量子fi阳离子的序列。SPECIfic Aim 2将检查 在当今典型的范围内，由于扩散时间的变化而导致的ff变化 临床磁共振系统。SPECIfic Aim 3引入了低噪声扩散数据的高级处理，它将 使用非侵入性和经济的外部线圈实现高精度和精密度是必不可少的。 SPECIfic Aim 4提出了一种新的模数转换计算方法，该方法能够完全捕捉复杂的ff效应。 信号在组织中衰减，同时在很大程度上与协议和系统无关。此外，重新- 根据该方法处理的测量图像和定量地图表现出相当低的噪声， 这可以换取更高的空间分辨率或更短的扫描持续时间。总而言之，康奈尔- 数据的量化性质及其无处不在的有效性和可比性将极大地促进 制定与疾病相关的门槛建议。最终，这可能会允许 更可靠地区分侵袭性疾病和惰性疾病。

Abstract Non-invasive, multi-parametric characterization of prostate cancer (PC), with magnetic resonance imaging (MRI) methods, is an active area of research with great potential for providing improved diagnosis and treatment monitoring. The PI-RADSv2 assessment system that was established by an international team of experts recognizes the value of quantitative images in PC diagnosis, but relies largely on qualitative evaluation of weighted images. Although this grading approach achieves reasonably good separation be- tween normal and abnormal prostate tissue, it does not achieve adequate separation between indolent and aggressive disease, with the risk that more unnecessary and costly surgery is performed with poten- tially dire consequences on patient quality of life. High-value protocols, without need for an invasive and costly endo-rectal radio-frequency coil are being investigated. This comes at the cost of extended scan time and reduced image quality in terms of spatial resolution, signal-to-noise ratio and signal bias, which negatively impacts sensitivity and specificity of multi-parametric MRI. With the pronounced increase of multi-parametric MRI exams, there is also the desire to integrate the support by the most recent revolution in diagnostic imaging, i.e., machine learning. It becomes increasingly clear, that in order to avoid having to train neural networks for each specific system and protocol, reproducible and thus preferably quantitative imaging protocols are essential. To overcome these limitations, we propose both pulse sequence develop- ment, investigation of ADC validity and reproducibility and novel post-processing strategies. The overall objective is to demonstrate the added value of lesion characterization with quantitative values and at the same time understand and minimize the influence of protocol choices and scan hardware, hence improve overall reproducibility. Specific Aim 1 will focus on the development of a low distortion MR imaging sequence for rapid concurrent quantification of T2 and diffusion signal decay. Specific Aim 2 will examine ADC variations that result from changes in diffusion time over a range that is typical with present day clinical MR systems. Specific Aim 3 introduces advanced handling of low noise diffusion data, which will be indispensable for achieving high accuracy and precision with non-invasive and economic external coils. Specific Aim 4 introduces a novel ADC computation approach that fully captures the complex diffusion signal decays in tissues and at the same time is largely protocol and system independent. Moreover, re- sulting images and quantitative maps processed according this approach, exhibit considerably lower noise, which can be traded for higher spatial resolution or shorter scan duration. In combination, the consis- tently quantitative nature of the data and its ubiquitous validity and comparability will greatly facilitate the establishment of recommendations for disease-related thresholds. Ultimately this may permit much more reliable differentiation of aggressive from indolent disease.