深度学习重建算法在肺核中的初步经验:超低剂量方案下对肺结节的临床实用性
本研究纳入68例同时接受胸部非增强超低剂量CT(ULD CT)和增强CT(CE CT)的患者,比较6种算法重建的ULD CT图像质量与结节锐利度,以CE CT肺核重建(CE AR50 Lung)为参考标准。结果显示TF Lung各强度均显著降低图像噪声(均p<0.001),低、中强度TF Lung的结节锐利度与参考标准相似。作者认为TF Lung有潜力提升ULD CT的图像质量和结节评估。
要点
- TF Lung各强度较AR50 Lung显著降低图像噪声(均p<0.001)。
- TF Lung低、中强度结节锐利度与参考标准CE AR50 Lung相似。
- TF Lung有潜力在不损害结节锐利度的情况下降低ULD CT图像噪声。
核心结论
- TF Lung有潜力提升超低剂量CT的图像质量和结节评估,低、中强度TF Lung的结节锐利度与参考标准相似。
研究设计与样本
- 研究纳入同时接受胸部非增强ULD CT和CE CT的患者68例,ULD CT采用6种算法重建(AR50 STD、AR50 Lung、TF标准核、TF Lung低/中/高强度),CE CT采用AR50 STD和AR50 Lung重建,共8组重建图像;以CE AR50 Lung为参考标准。
关键影像表现/征象
- TF Lung各强度重建较AR50 Lung显著降低图像噪声(均p<0.001)。
- TF Lung低、中强度重建的结节锐利度与参考标准CE AR50 Lung相似。
- 对病理确诊恶性结节的恶性相关影像特征进行了评估。
诊断标准与关键数值
- 结节锐利度通过计算半高全宽(full-width half-maximum)值评估。
- 客观图像质量比较指标包括图像噪声、信噪比和对比噪声比。
- 以CE CT肺核重建(CE AR50 Lung)作为参考标准。
临床应用与阅片要点
- 在超低剂量CT肺结节评估中,TF Lung低、中强度可在降低图像噪声的同时保持与参考标准相似的结节锐利度。
- TF Lung可能支持放射科医生对恶性结节的诊断决策。
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
OBJECTIVES: TrueFidelity (TF), a deep learning image reconstruction algorithm that was originally available only in standard kernel, has recently become available in lung kernel. This is the first study to assess TF in lung kernel (TF Lung) on image quality and nodule sharpness at ultra-low-dose CT (ULD CT).
MATERIALS AND METHODS: This study included patients who underwent both non-contrast ULD CT and contrast-enhanced CT (CE CT) of the chest. ULD CT scans were reconstructed using 6 algorithms, including Adaptive statistical iterative reconstruction-V in standard kernel (AR50 STD), AR50 in lung kernel (AR50 Lung), TF in standard kernel, and TF Lung at low, medium, and high strengths. CE CT scans were reconstructed using AR50 STD and AR50 Lung. In total, 8 sets of reconstruction images were obtained and reviewed for each patient. Objective image quality, such as image noise, signal-to-noise ratio, and contrast-to-noise ratio, was compared. Nodule sharpness was evaluated by calculating the full-width half-maximum value. Malignancy-related imaging features in nodules that had been pathologically confirmed as malignant were evaluated. CE CT reconstructed in the lung kernel (CE AR50 Lung) served as the reference standard.
RESULTS: A total of 68 patients underwent analysis. ULD CT reconstructed with TF Lung at all strengths significantly decreased image noise compared to AR50 Lung (all p < 0.001). TF Lung in low and medium strengths showed similar nodule sharpness compared to the reference standard.
CONCLUSION: TF Lung has the potential to enhance both image quality and nodule evaluation in ULD CT scans.
KEY POINTS: Question How does TrueFidelity in the lung kernel (TF Lung) perform in terms of image quality and lung nodule assessment in ultra-low-dose (ULD) CT scans? Findings TF Lung significantly decreased image noise, and TF Lung in low and medium strengths showed similar nodule sharpness compared to the reference standard. Clinical relevance TF Lung has the potential to reduce image noise without compromising the sharpness of nodules in ULD CT scans. TF Lung may support radiologists' diagnostic decision-making for malignancy.
引用
- Kim J, Kang S, Hwang MH, et al. Initial experience of deep learning reconstruction algorithm in lung kernel: clinical usefulness for lung nodules at ultra-low-dose protocol. Eur Radiol. 2026. PMID: 41817709. doi:10.1007/s00330-026-12404-y
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。