MRI影像组学预测脑转移瘤生存:机器学习分析
该回顾性研究在198例公共脑转移瘤队列中构建T1增强MRI影像组学生存模型,并在69例独立队列中外部验证。最终模型保留7个非零特征,训练队列C-index为0.615,外部验证为0.574;影像组学联合GPA模型C-index为0.591。结论认为T1增强MRI影像组学单独区分总生存能力有限,应谨慎作为探索性生物标志物。
要点
- T1增强MRI影像组学单独预测脑转移瘤总生存的区分能力有限,外部验证C-index仅0.574。
- 影像组学联合GPA模型C-index为0.591,提示需整合临床等多变量构建预后模型。
- 模型基于三个最大病灶聚合的患者级特征,采用弹性网Cox回归和10折交叉验证。
核心结论
- T1增强MRI影像组学对脑转移瘤患者总生存的单独区分能力有限,支持其作为探索性影像生物标志物谨慎使用。
- 研究强调需要整合临床、治疗、分子和系统性疾病变量的综合预后模型。
研究设计与样本
- 回顾性研究,在198例公共脑转移瘤队列中开发T1增强MRI影像组学生存模型,并在69例独立队列中外部验证固定模型(不重新拟合或校准)。
- 患者级影像组学特征由三个最大病灶的病灶级特征聚合得到,采用弹性网Cox模型并以10折交叉验证选择。
- 模型区分度用Harrell C-index评估,95%CI基于2000次bootstrap重采样。
诊断标准与关键数值
- 最终模型经相关性过滤和弹性网选择后保留7个非零T1增强影像组学特征。
- 影像组学评分在训练队列的C-index为0.615(95%CI 0.543-0.687),外部验证队列为0.574(95%CI 0.491-0.658)。
- 在具有GPA的外部验证子集中,影像组学联合GPA模型的C-index为0.591(95%CI 0.509-0.672)。
临床应用与阅片要点
- 日常阅片和报告时可关注T1增强MRI影像组学特征作为探索性预后信息,但不宜单独用于生存预测决策。
- 报告预后相关影像分析时应结合GPA等临床预后因素,避免孤立解读影像组学评分。
局限性
- 回顾性研究设计。
- 影像组学模型单独区分能力有限,外部验证C-index较低。
- 研究未整合临床、治疗、分子和系统性疾病变量于最终影像组学模型。
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: Brain metastases carry poor prognosis, and accurate survival prediction is critical for treatment planning. Radiomics offers a means of extracting high-dimensional imaging biomarkers, but its prognostic utility in brain metastasis remains unclear.
PURPOSE: To evaluate whether MRI-derived radiomic features can improve survival prediction in patients with brain metastases.
MATERIALS AND METHODS: This retrospective study developed a T1 postcontrast MRI radiomics survival model in a public brain metastasis cohort of 198 patients and externally validated the fixed radiomics model, without refitting or recalibration, in an independent cohort of 69 patients. Patient-level radiomics features were derived from lesion-level features aggregated across the three largest lesions. An elastic-net Cox model was selected with 10-fold cross-validation. Model discrimination was assessed with Harrell C-index and 2000 bootstrap resamples for 95% CIs.
RESULTS: The final model retained 7 nonzero T1 postcontrast radiomics features after correlation filtering and elastic-net selection. The radiomics score had a C-index of 0.615 (95% CI, 0.543-0.687) in the training cohort and 0.574 (95% CI, 0.491-0.658) in external validation. In the external-validation subset with available Graded Prognostic Assessment, the combined radiomics plus Graded Prognostic Assessment model had a C-index of 0.591 (95% CI, 0.509-0.672).
CONCLUSION: T1 postcontrast MRI radiomics showed limited standalone discrimination for overall survival in patients with brain metastases. These results support cautious use of radiomics as an exploratory imaging biomarker and emphasize the need for integrated prognostic models that include clinical, treatment, molecular, and systemic disease variables.
引用
- Salim HA, Calabrese E, Naeem A, et al. MRI Radiomics for Survival Prediction in Brain Metastases: A Machine Learning Analysis. AJNR Am J Neuroradiol. 2026. PMID: 42716711. doi:10.3174/ajnr.A9630
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。