深度学习加速单次屏气腹部HASTE序列:3T可变重聚焦翻转角开发与评估
本研究在3T上开发可变翻转角演化的深度学习加速单次屏气HASTE序列,用于腹部T2加权脂肪抑制成像。志愿者研究确定最优翻转角演化为130-90-110-130度,51例患者临床评估显示图像质量优秀,与常规BLADE序列检测的肝脏病灶数目和直径无差异,采集时间仅21秒(BLADE为4分钟)。
要点
- HASTEDL序列可在3T且SAR限制内实现单次屏气21秒采集,图像质量中位评分为4分。
- HASTEDL与BLADE检测的肝脏病灶数目(n=102)和测量直径无差异。
- BLADE在整体图像质量、噪声、对比度、锐利度、伪影、心脏运动相关信号丢失及诊断信心上评分显著更高。
关键影像表现/征象
- HASTEDL采用翻转角演化控制点130-90-110-130度,在5种演化中被选为最优。
- HASTEDL图像质量中位评分为4分(四分位距3-4),所有51例患者均成功采集。
- BLADE在整体图像质量、噪声、对比度、锐利度、伪影、心脏运动相关信号丢失及诊断信心上评分显著高于HASTEDL。
- 两种序列检测的肝脏病灶数目均为102个,测量直径无差异。
诊断标准与关键数值
- 图像质量等采用Likert 1-4分量表评估,4分为最佳。
- HASTEDL采集时间为21秒(单次屏气),BLADE采集时间为4分钟(呼吸触发)。
临床应用与阅片要点
- HASTEDL可作为3T腹部T2加权脂肪抑制成像的快速替代序列,显著缩短屏气时间。
- 尽管BLADE主观评分更高,但HASTEDL在肝脏病灶检测数目和直径上与BLADE无差异,可满足诊断需求。
- 需注意HASTEDL在噪声、对比度、锐利度、伪影及心脏运动相关信号丢失方面主观评分低于BLADE。
局限性
- 研究为单中心前瞻性设计,样本量有限(志愿者11例,患者51例)。
- 仅与BLADE序列比较,未涉及其他T2加权序列。
- 主观评分采用Likert量表,未报告读者间一致性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
OBJECTIVE: Deep learning (DL) reconstruction enables substantial acceleration of image acquisition while maintaining diagnostic image quality. The aims of this study were to overcome the drawback of specific absorption rate (SAR)-related limitations at 3 T and to develop a DL-accelerated single-breath-hold half-Fourier acquisition single-shot turbo spin echo (HASTE) sequence for 2-dimesional T2-weighted fat-suppressed magnetic resonance imaging of the abdomen at 3 T using a variable flip angle (FA) evolution for the refocusing radiofrequency pulses, as well as to evaluate its feasibility and image quality in comparison to state-of-the-art T2-weighted fat-suppressed imaging technique (BLADE).
MATERIALS AND METHODS: First, a suitable FA evolution with low cardiac motion-related signal loss (CRSL) and low SAR was determined through a prospective volunteer study with 11 participants. Image quality and diagnostic confidence with 5 different FA evolutions of a HASTEDL were assessed to identify the most suitable FA evolution. Second, the identified FA evolution was implemented clinically and evaluated in 51 patients undergoing a clinically indicated liver magnetic resonance imaging at 3 T. Two radiologists assessed the HASTEDL and standard sequences regarding overall image quality, noise, contrast, sharpness, artifacts, CRSL, and diagnostic confidence using a Likert scale ranging from 1 to 4, with 4 being the best. Comparative analyses were conducted to assess the differences between HASTEDL (acquisition time, 21 seconds; single breath-hold) and the routinely used T2-weighted BLADE sequence (acquisition time, 4 minutes; respiratory triggering).
RESULTS: From the volunteer study, the FA evolution characterized by the control points 130-90-110-130 degrees (HASTEDL) was identified as optimal among the 5 evolutions evaluated and was implemented in our clinical protocol. In all 51 patients, HASTEDL was successfully acquired at 3 T and showed excellent image quality (median, 4; interquartile range, 3-4). Although BLADE was rated significantly higher for overall image quality, noise, contrast, sharpness, artifacts, CRSL, and diagnostic confidence than HASTEDL, no differences were found concerning the number (n = 102) and measured diameter of the detected hepatic lesions between the 2 sequences BLADE and HASTEDL.
CONCLUSIONS: The proposed single-breath-hold abdominal HASTEDL with variable refocusing FAs is feasible at 3 T within SAR limits and yields high image quality and diagnostic confidence as compared with a standard T2-weighted acquisition technique, at a 10th of the acquisition time.
引用
- Herrmann J, Nickel D, Mugler JP, et al. Development and Evaluation of Deep Learning-Accelerated Single-Breath-Hold Abdominal HASTE at 3 T Using Variable Refocusing Flip Angles. Invest Radiol. 2021. PMID: 33965966. doi:10.1097/RLI.0000000000000785
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。