直方图分析在平扫CT鉴别高密度肾囊肿与实性肾肿块中的增量价值
回顾性分析99个高密度囊性病变与28个实性恶性病变,评估体素直方图分析在平扫CT上的鉴别价值。直方图分析(第75/90百分位±1标准差)诊断准确率分别为77.2%和68.5%,高于70 HU均值阈值的37.0%。结论提示直方图分析可能优于70 HU标准,但需多机构研究外部验证。
要点
- 直方图分析诊断准确率77.2%(第75百分位±1SD)和68.5%(第90百分位±1SD),高于70 HU均值阈值的37.0%。
- 88.9%病变的体素衰减值呈高斯分布,支持从单次测量计算直方图参数。
- 第75百分位阈值≥65 HU可能在不降低特异度的情况下提高高密度肾囊肿诊断敏感度。
核心结论
- 直方图分析利用同一感兴趣区测量病变衰减,可能为平扫CT高密度囊肿表征提供超越70 HU标准的额外价值,并可能影响临床决策;需多机构研究外部验证以确认其普遍性和更广泛适用性。
研究设计与样本
- 回顾性分析,纳入99个高密度囊性病变和28个实性恶性病变,数据来自放射学数据库(2015年至2021年)和病理学数据库(2010年至2020年);在病变中心绘制感兴趣区后分析体素衰减值分布;参考标准为恶性肿块的组织病理学诊断和囊肿的增强CT或MRI。
诊断标准与关键数值
- 第75百分位±1标准差准确率77.2%(95%CI 68.9%-84.2%),第90百分位±1标准差准确率68.5%(95%CI 59.7%-76.4%),70 HU阈值标准准确率37.0%(95%CI 28.6%-46.0%);第75百分位阈值≥65 HU可能提高高密度囊肿诊断敏感度而不降低特异度。
临床应用与阅片要点
- 作者认为直方图分析可能为平扫CT高密度囊肿表征提供超越70 HU标准的额外价值并可能影响临床决策;第75百分位阈值≥65 HU可能提高诊断敏感度而不降低特异度;需多机构研究外部验证。
局限性
- 需多机构研究进行外部验证以确认普遍性和更广泛适用性。
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: To investigate the utility of voxel histogram analysis (HA) for differentiating hyperdense renal cysts from small solid masses on unenhanced CT scans.
METHODS: A retrospective analysis of 99 hyperdense cystic lesions and 28 solid malignant lesions was conducted using a radiological database (from 2015 to 2021) and a pathological database (from 2010 to 2020). The study investigated the distribution of voxel attenuation values using percentiles to establish reliable criteria for differentiation after drawing a region of interest (ROI) in the centre of the lesions. The standard of reference was a histopathological diagnosis for malignant masses and contrast-enhanced CT or MRI for cysts.
RESULTS: HA provided higher diagnostic accuracy than the conventional mean attenuation value of 70 Hounsfield Units (HU). For the 75th and 90th percentiles ± 1 standard deviation, accuracies of 77.2% (95% confidence interval 68.9%-84.2%) for the 75th and 68.5% (59.7%-76.4%) for the 90th were found, versus 37.0% (28.6%-46.0%) for the 70 HU threshold criterion. A Gaussian distribution of voxel attenuation values was observed in 88.9% of the lesions, suggesting that it is feasible to calculate these parameters from a single measurement.
CONCLUSION: The study underscores the potential of HA as a valuable tool for characterizing hyperdense cysts on unenhanced CT by using the same ROI for measuring lesion attenuation. HA could offer additional value beyond the 70 HU criterion and possibly influence clinical decisions. Multi-institutional studies are necessary for external validation to confirm its generalizability and more extensive applicability.
ADVANCES IN KNOWLEDGE: (1) A single measurement on unenhanced CT images, using mean attenuation and standard deviation, accurately reflects the voxel distribution of both cystic and solid masses, allowing for the application of histogram analysis. (2) The 75th percentile threshold of 65 HU or higher could potentially increase sensitivity in diagnosing hyperdense cysts, compared to the 70 HU mean attenuation threshold, without compromising specificity.
引用
- de Souza Torres CV, de Freitas Secaf A, Maia Vieira DF, et al. The incremental value of histogram analysis in the differentiation between hyperdense cysts and solid renal masses on unenhanced CT images. Br J Radiol. 2025. PMID: 39383171. doi:10.1093/bjr/tqae198
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。