深度学习重建用于肝脏DWI：对图像质量和ADC定量的影响
该回顾性研究纳入188例患者，比较压缩感知（CS）与基于模型的深度学习重建（DL-DWI）在肝脏DWI中的图像质量、ADC值及病灶检出。结果显示DL-DWI在信噪比、对比噪声比、图像质量评分及病灶显著性上均优于CS-DWI，尤其对小病灶和浅表病灶。
要点
- DL-DWI可提高肝脏DWI的信噪比和对比噪声比，改善图像质量评分。
- DL-DWI对小于10mm及距肝表面1cm内的病灶显著性更高。
- DL-DWI的ADC值显著低于CS-DWI，报告ADC时需注意重建方法影响。
核心结论
- 基于模型的DL重建改善了肝脏DWI的定性和定量图像质量及病灶显著性，尤其对小和浅表病灶。
研究设计与样本
- 回顾性分析188例（男103例，女85例，平均年龄69.6岁）于2024年10月至2025年1月行肝脏DWI的患者。
- DL-DWI采用Adaptive-CS-Net，一种嵌入CS重建流程的基于模型的深度学习架构。
- 两名放射科医生独立进行定量和定性评估，包括病灶显著性评估。
关键影像表现/征象
- DL-DWI所有器官的信噪比均高于CS-DWI（p < 0.05）。
- DL-DWI所有器官的变异系数（除左外侧段和椎旁肌外）和ADC值均显著低于CS-DWI（p < 0.05）。
- 所有病灶（肝细胞癌、转移瘤和血管瘤；n = 155）的对比噪声比在DL-DWI中显著大于CS-DWI（p < 0.05）。
- 肝脏轮廓、肝脏变形、图像噪声、伪影和整体图像质量评分在DL-DWI中显著高于CS-DWI（p < 0.05）。
- 整体病灶显著性（n = 227）在DL-DWI中显著高于CS-DWI（p < 0.001）；小病灶（<10 mm，n = 72）和浅表病灶（距肝表面1 cm内，n = 75）的显著性也显著更高（p < 0.05）。
临床应用与阅片要点
- DL-DWI可改善图像质量和病灶显著性，尤其适用于评估小病灶和浅表病灶。
- 由于DL-DWI的ADC值显著低于CS-DWI，临床解读ADC时应考虑重建方法的影响。
局限性
- 回顾性研究设计。
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
PURPOSE: To assess the image quality, apparent diffusion coefficient (ADC) values, and lesion detection of diffusion-weighted imaging (DWI) with compressed sensing (CS) and model-based deep learning (DL) reconstruction (DL-DWI) by comparing with CS-DWI.
MATERIALS AND METHODS: We retrospectively analyzed 188 patients (103 men and 85 women; mean age, 69.6 years) who underwent liver DWI between October 2024 and January 2025. DL-DWI incorporated Adaptive-CS-Net, a model-based deep learning architecture embedded within the CS reconstruction pipeline. Two radiologists independently performed quantitative and qualitative evaluations, including lesion conspicuity assessment.
RESULTS: Signal-to-noise ratio in DL-DWI was higher than that in CS-DWI for all organs (p < 0.05). Coefficient of variation (except for lateral segment and paraspinal muscle) and ADC values in DL-DWI were significantly lower than those in CS-DWI for all organs (p < 0.05). Contrast-to-noise ratio of all lesions (hepatocellular carcinoma, metastasis, and hemangioma; n = 155) was significantly larger in DL-DWI than in CS-DWI (p < 0.05). Image quality scores for the liver contours, distortion of the liver, image noise, artifacts, and overall image quality were significantly higher for DL-DWI than for CS-DWI (p < 0.05). Overall lesion conspicuity (n = 227) was significantly higher in DL-DWI than in CS-DWI (p < 0.001). Lesion conspicuity of small lesions (<10 mm, n = 72) and superficial lesions (within 1 cm from the liver surface, n = 75) in DL-DWI was significantly higher than in CS-DWI (p < 0.05).
CONCLUSION: Model-based DL reconstruction improved qualitative and quantitative image quality and lesion conspicuity, particularly for small and superficial lesions.
引用
- Ozaki K, Hasegawa H, Ishida S, et al. Deep learning reconstruction for liver DWI: impact on image quality and ADC quantification. Eur J Radiol. 2026. PMID: 42134117. doi:10.1016/j.ejrad.2026.112925
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。