深度学习图像重建算法在亚毫西弗腹部盆腔CT中检测尿路结石
57例疑似尿路结石患者同时接受低剂量(LD)和超低剂量(ULD)腹部盆腔CT,ULD原始数据分别用ASIR-V 70%和DLIR-H重建,以LD CT联合ASIR-V 70%为参考标准。38例共检出266枚结石,ULD CT联合DLIR-H的敏感度和特异度分别为89%和94%,ULD CT有效剂量0.59 mSv低于LD CT的1.96 mSv。DLIR-H图像噪声显著低于ASIR-V 70%,信噪比和主观图像质量显著更高。
要点
- ULD CT联合DLIR-H检测尿路结石的敏感度为89%、特异度为94%。
- ULD CT平均有效剂量0.59 mSv,LD CT为1.96 mSv。
- DLIR-H图像噪声显著低于ASIR-V 70%,信噪比显著更高。
核心结论
- ULD CT联合深度学习图像重建在尿路结石中保持良好的诊断性能,图像质量优于混合迭代重建,且辐射剂量显著降低。
研究设计与样本
- 57例疑似尿路结石患者接受非增强低剂量(LD)和超低剂量(ULD)腹部盆腔CT;ULD原始数据用ASIR-V 70%和DLIR-H重建,以LD CT联合ASIR-V 70%为参考标准,由两名阅片者评估。
诊断标准与关键数值
- ULD CT联合DLIR-H检测尿路结石的敏感度89%、特异度94%。
- 两名阅片者ULD CT联合DLIR-H与LD CT联合ASIR-V 70%的观察者内一致性ICC均为0.90。
- ULD CT平均有效剂量0.59 mSv,LD CT平均有效剂量1.96 mSv。
临床应用与阅片要点
- DLIR-H重建的ULD CT图像噪声显著低于ASIR-V 70%,信噪比显著更高,主观图像质量显著更好。
局限性
- 原文未报告局限性。
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
PURPOSE: Urolithiasis is a chronic condition that leads to repeated CT scans throughout the patient's life. The goal was to assess the diagnostic performance and image quality of submillisievert abdominopelvic computed tomography (CT) using deep learning-based image reconstruction (DLIR) in urolithiasis.
METHODS: 57 patients with suspected urolithiasis underwent both non-contrast low-dose (LD) and ULD abdominopelvic CT. Raw image data of ULD CT were reconstructed using hybrid iterative reconstruction (ASIR-V 70%) and high-strength-level DLIR (DLIR-H). The performance of ULD CT for the detection of urinary stones was assessed by two readers and compared with LD CT with ASIR-V 70% as a reference standard. Image quality was assessed subjectively and objectively.
RESULTS: 266 stones were detected in 38 patients. Mean effective dose was 0.59 mSv for ULD CT and 1.96 mSv for LD CT. For diagnostic performance, sensitivity and specificity were 89% and 94%, respectively, for ULDCT with DLIR-H. There was an almost perfect intra-observer concordance on ULD CT with DLIR-H versus LDCT with ASIR-V 70% (ICC = 0.90 and 0.90 for the two readers). Image noise was significantly lower and signal-to-noise ratio significantly higher with DLIR-H compared to ASIR-V 70%. Subjective image quality was also significantly better with ULDCT with DLIR-H.
CONCLUSION: ULD CT with Deep Learning Image Reconstruction maintains a good diagnostic performance in urolithiasis, with better image quality than hybrid iterative reconstruction and a significant radiation dose reduction.
引用
- Prod'homme S, Bouzerar R, Forzini T, et al. Detection of urinary tract stones on submillisievert abdominopelvic CT imaging with deep-learning image reconstruction algorithm (DLIR). Abdom Radiol (NY). 2024. PMID: 38470506. doi:10.1007/s00261-024-04223-w
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。