单次屏气T2加权肝脏成像改善病灶定性与LI-RADS分级
该前瞻性研究纳入141例慢性肝病患者,比较自由呼吸TSE、单次激发TSE及联合深度学习重建三种肝脏T2加权序列。结果显示单次激发TSE显著减少良性病灶被误判为恶性倾向及HCC T2特征不可识别,LI-RADS LR-3升级至LR-4/5比例明显提高,采集时间缩短91.8%。
要点
- 单次激发T2加权较自由呼吸TSE显著提高HCC的LI-RADS分级,LR-3升级比例从40.4%提升至67.1%。
- 单次激发TSE可减少良性病灶的「倾向恶性」误判和HCC的T2特征不可识别。
- 联合深度学习重建不改变LI-RADS结局,但图像质量和病灶显著度更好。
核心结论
- 单次激发TSE独立于深度学习重建,在肝脏T2加权辅助征象评估和HCC的LI-RADS分级上优于自由呼吸TSE。
- 深度学习重建不改变诊断分级结果,但改善图像质量和病灶显著度。
- 单次激发TSE将采集时间从164秒缩短至13.5秒,减少91.8%。
研究设计与样本
- 前瞻性研究,纳入141例慢性肝病患者,行肝脏MRI含三种T2加权序列:自由呼吸TSE(164秒)、单次激发TSE(13.5秒)、单次激发TSE联合深度学习重建(13.5秒)。
- 定量分析127个≥10 mm病灶的信噪比、信号均匀性和病灶-肝脏对比噪声比。
- 两名放射科医师独立评估340个≥5 mm病灶的定性图像参数和LI-RADS分级,其中213个HCC、104个囊肿、23个血管瘤。
关键影像表现/征象
- 单次激发TSE联合深度学习重建在信噪比、信号均匀性、病灶-肝脏对比噪声比方面均显著高于自由呼吸TSE和单纯单次激发TSE,p均≤0.032。
- 肝脏边缘锐利度、血管显示、伪影抑制和整体图像质量评分均以单次激发TSE联合深度学习重建最优,p均≤0.046。
- 与自由呼吸TSE相比,单次激发TSE及联合深度学习重建显著减少良性病变的「倾向恶性」误判和HCC的T2「不可识别」特征。
诊断标准与关键数值
- LI-RADS升级:单次激发TSE与联合深度学习重建使LR-3升级至LR-4/5的比例分别从40.4%和41.8%提升至67.1%和68.1%,p<0.001。
- 整体LI-RADS分级分布变化具有统计学意义,p=0.018。
- 采集时间缩短:从自由呼吸TSE的164秒降至单次激发序列的13.5秒,减少91.8%。
临床应用与阅片要点
- 肝脏MRI的T2加权可优先采用单次激发TSE序列,有助于提高HCC的LI-RADS分级准确性。
- 单次激发TSE可减少良性病灶被过度判为恶性倾向,降低假阳性风险。
- 深度学习重建可作为提升图像质量的选择,但不影响LI-RADS诊断分级。
局限性
- 原文未报告研究局限。
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
RATIONALE AND OBJECTIVES: T2-weighted imaging is essential for liver magnetic resonance imaging (MRI), but conventional free-breathing turbo spin-echo (FB-TSE) is prolonged and motion-sensitive. Single-shot techniques offer motion robustness but compromise image quality. The aim is to evaluate whether single-shot turbo spin-echo (SS-TSE) provides superior lesion conspicuity and accurate Liver Imaging Reporting and Data System (LI-RADS) categorization compared with FB-TSE, and whether deep learning-based reconstruction (SS-TSE+DLR) preserves this benefit while improving image quality.
MATERIALS AND METHODS: This prospective study enrolled 141 patients with chronic liver disease who underwent liver MRI with three T2-weighted sequences: FB-TSE (164 s), SS-TSE (13.5 s), and SS-TSE+DLR (13.5 s). Quantitative analysis assessed signal-to-noise ratio (SNR), signal homogeneity, and lesion-to-liver contrast-to-noise ratio (CNR) for 127 lesions ≥10 mm. Two radiologists independently evaluated qualitative image parameters and LI-RADS categories for 340 lesions ≥5 mm, including 213 hepatocellular carcinomas (HCCs), 104 cysts, and 23 hemangiomas.
RESULTS: SS-TSE+DLR demonstrated significantly higher SNR, signal homogeneity, and CNR than both FB-TSE and SS-TSE (all p≤0.032). Qualitative scores for liver edge sharpness, vessel delineation, artifact suppression, and overall image quality favored SS-TSE+DLR (all p ≤ 0.046). Compared with FB-TSE, SS-TSE and SS-TSE+DLR significantly reduced false "favoring malignancy" assignments in benign lesions and "not identifiable" T2 features in HCC, facilitating significant upgrading from LR-3 to LR-4/5 (from 40.4%/41.8% to 67.1%/68.1%; p < 0.001) with significant overall LI-RADS categorization shift (p = 0.018).
CONCLUSION: SS-TSE, independent of DLR, improved T2-weighted ancillary feature assessment and LI-RADS categorization of HCC over FB-TSE via a 91.8% reduction in acquisition time; DLR did not alter these outcomes but improved image quality and lesion conspicuity.
引用
- Ozaki, Haseyama, Sugioka, et al. Single-shot T2-Weighted Turbo Spin-echo Imaging of the Liver Improves Lesion Characterization Compared with Free-breathing Acquisition. Acad Radiol. 2026. PMID: 42629290. doi:10.1016/j.acra.2026.08.023
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。