Development of Magnetic Resonance Fingerprinting in Kidney for Evaluation of Renal Cell Carcinoma

摘要 预计到2021年，肾癌将影响美国76，080名新患者，其中13，780人死亡。肾 细胞癌（RCC）是最常见的肾癌类型，其造成显著的经济负担 在医疗保健系统上。最近一项基于SEER医疗保险数据库的研究报告称， 每个RCC患者的成本为23，489美元，加权总经济负担为21亿美元。肾细胞癌常表现为 偶然发现的不完全特征性的肾脏肿块。这些患者中有许多人患有偶发性肾功能衰竭， 肿块直接手术或活检，无需进一步影像学评估， 用目前的成像技术诊断并不总是可能的。然而，前期手术或活检不是 理想情况下，近25%的偶发性肾脏肿块为良性（血管平滑肌脂肪瘤、嗜酸细胞瘤）或低度恶性 （嫌色细胞肾细胞癌，低级别透明细胞肾细胞癌）和过度治疗这样的群众增加了不必要的 发病率和卫生保健费用。先前的研究表明，低级别RCC可以保守治疗， 在选定的患者（老年患者和不适合手术的患者）中进行积极监测，但在 目前还没有一种非侵入性的方法来区分低级别RCC和侵袭性RCC（高级别透明细胞 RCC，乳头状RCC）。因此，迫切需要开发新的非侵入性定量分析方法。 用于准确表征肾脏肿块的生物标志物，以便更多患者符合主动监测的条件 可以被识别。最近的研究表明，包括T1、T2和T2* 在内的MR组织弛豫测量图 映射和脂肪分数定量可以提供肾脏疾病的改进表征， 与肾癌的肿瘤分级和生物学侵袭性有关。然而，目前的肾舒张测量映射 这些技术仍然受到长时间屏气、有限的空间分辨率/覆盖范围以及大部分捕获能力的限制 一次一个组织属性。此外，定量测量通常容易受到运动伪影的影响， 重复性和再现性差。在这项研究中，我们建议利用新的MR指纹（MRF） 技术与机器学习方法一起来减轻肾脏成像中的上述限制。在 特别是，我们将开发一种新的3D自由呼吸肾脏MRF方法，用于同时测量T1，T2，T2* 和脂肪 分数定量（目标1）。我们将联合收割机将这种肾脏MRF采集与新颖的深度学习相结合 加速数据采集和提高组织标测效率的方法（目标2）。最后，我们将应用 MRF技术在RCC患者中的应用，以探索其在表征肾癌方面的诊断强度（目的 3）。成功开发后，MRF获得的多参数定量测量可以使 MRI是诊断和预测肾细胞癌分级的有力工具，其最终目的是 在符合条件的良性/低级别RCC患者中消除不必要的活检/手术，并提供指导 找到最合适的治疗策略

Abstract Kidney cancer is expected to affect 76,080 new patients with 13,780 deaths in the U.S. in the year 2021. Renal cell carcinoma (RCC) is the most common type of kidney cancer which imposes significant economic burden on healthcare system. A recent study based on SEER Medicare database reported that the total healthcare cost per RCC patient was $23,489 with a weighted total economic burden of $2.1 billion. RCC often presents as an incidentally detected, incompletely characterized renal mass. Many of these patients with incidental renal mass either undergo direct surgery or biopsy without further imaging evaluation as accurate histologic diagnosis with current imaging techniques is not always possible. However, upfront surgery or biopsy is not ideal as nearly 25% incidental renal masses are either benign (angiomyolipoma, oncocytoma) or low-grade (chromophobe RCC, low-grade clear cell RCC) and overtreatment of such masses adds to unnecessary morbidity and health care cost. Prior studies have shown low-grade RCC can be managed conservatively with active surveillance in select patients (elderly patients and patients who are poor surgical candidates), but at present there is a no non-invasive way to separate low-grade RCC from aggressive RCC (high-grade clear cell RCC, papillary RCC). Accordingly, there is an emergent need to develop novel non-invasive quantitative biomarkers for accurate characterization of renal masses so that more patients eligible for active surveillance could be identified. Recent studies have shown that MR tissue relaxometry mapping including T1, T2 and T2* mapping and fat fraction quantification can provide improved characterization of kidney diseases and correlate with tumor grade and biologic aggressiveness in RCC. However, the current kidney relaxometry mapping techniques still suffer from long breath-holds, limited spatial resolutions/coverage, and ability to mostly capture one tissue property at a time. Further, the quantitative measures are often susceptible to motion artifacts with poor repeatability and reproducibility. In this study, we propose to utilize the novel MR Fingerprinting (MRF) technique together with machine learning methods to mitigate aforementioned limitations in kidney imaging. In particular, we will develop a new 3D free-breathing kidney MRF method for simultaneous T1, T2, T2* and fat fraction quantification (Aim 1). We will combine this kidney MRF acquisition with novel deep learning approaches to accelerate data acquisition and improve tissue mapping efficiency (Aim 2). Finally, we will apply the MRF technique in patients with RCC to explore its diagnostic strength in characterizing kidney cancer (Aim 3). Upon successful development, the multi-parametric quantitative measures acquired with MRF could make MRI a more powerful tool for the diagnosis and predicting of tumor grade in RCC, with the ultimate goal to eliminate unnecessary biopsy/surgery in eligible patients with benign/low-grade RCCs and provide guidance towards the most appropriate treatment strategy.