脑转移瘤:大小分布的统计建模启示
研究对68例未经治疗的脑实质转移瘤患者的增强T1加权 spoiled 梯度回波MRI图像测量肿瘤直径并计算体积,将肿瘤体积按秩次分布与11种统计曲线类型比较。39例最符合幂律分布(平均R²=0.938±0.050),20例最符合指数分布(平均R²=0.957±0.050);似然比分析显示66/68例平均R值为正(1.596±1.316),偏向幂律分布。结论认为未经治疗脑转移瘤的大小分布 favored 幂律分布,提示转移瘤并非孤立存在而是复杂系统的一部分,可能存在少量潜在变量显著影响该系统行为。
要点
- 未经治疗脑转移瘤的体积分布多数(39/68)最符合幂律分布,平均R²=0.938±0.050。
- 20/68例最符合指数分布,平均R²=0.957±0.050。
- 似然比分析66/68例平均R值为正(1.596±1.316),偏向幂律分布。
关键影像表现/征象
- 肿瘤体积基于增强T1加权 spoiled 梯度回波图像测量的肿瘤直径计算。
- 39/68例最符合幂律分布,平均R²=0.938±0.050。
- 20/68例最符合指数分布,平均R²=0.957±0.050。
- 其余病例分散于其余分布类型。
- 似然比分析显示66/68例平均R值为正(1.596±1.316),偏向幂律分布。
局限性
- 原文在结论中提出这些结果 suggest 可能存在少量潜在变量显著影响该系统行为,识别这些变量可能对理解这些病变和治疗能力产生深远影响。
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
BACKGROUND AND PURPOSE: Brain metastases are a common finding on brain MRI. However, the factors that dictate their size and distribution are incompletely understood. Our aim was to discover a statistical model that can account for the size distribution of parenchymal metastases in the brain as measured on contrast-enhanced MR imaging.
MATERIALS AND METHODS: Tumor volumes were calculated on the basis of measured tumor diameters from contrast-enhanced T1-weighted spoiled gradient-echo images in 68 patients with untreated parenchymal metastatic disease. Tumor volumes were then placed in rank-order distributions and compared with 11 different statistical curve types. The resultant R 2 values to assess goodness of fit were calculated. The top 2 distributions were then compared using the likelihood ratio test, with resultant R values demonstrating the relative likelihood of these distributions accounting for the observed data.
RESULTS: Thirty-nine of 68 cases best fit a power distribution (mean R 2 = 0.938 ± 0.050), 20 cases best fit an exponential distribution (mean R 2 = 0.957 ± 0.050), and the remaining cases were scattered among the remaining distributions. Likelihood ratio analysis revealed that 66 of 68 cases had a positive mean R value (1.596 ± 1.316), skewing toward a power law distribution.
CONCLUSIONS: The size distributions of untreated brain metastases favor a power law distribution. This finding suggests that metastases do not exist in isolation, but rather as part of a complex system. Furthermore, these results suggest that there may be a relatively small number of underlying variables that substantially influence the behavior of these systems. The identification of these variables could have a profound effect on our understanding of these lesions and our ability to treat them.
引用
- Buller M, Chapple KM, Bird CR, et al. Brain Metastases: Insights from Statistical Modeling of Size Distribution. AJNR Am J Neuroradiol. 2020. PMID: 32241770. doi:10.3174/ajnr.A6496
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。