子宫内膜癌:表观扩散系数图的纹理分析及其与组织病理学表现和预后的相关性
本研究纳入121例子宫内膜癌患者,采用1.5-T MRI及b值0和1000 sec/mm²的扩散加权成像,在ADC图上提取45个纹理特征(含高阶纹理特征),用随机森林模型与ADC值比较预测组织学分级和复发。随机森林模型预测高级别子宫内膜癌的AUC显著大于ADC值(0.967 vs 0.898,P=0.0336);预测复发的AUC两者差异无统计学意义(0.890 vs 0.875,P=0.7248)。结论认为ADC图纹理分析预测组织学分级的诊断性能显著高于ADC值,且对无复发生存期的预测比组织学分级和ADC值更有用。
要点
- 随机森林模型预测高级别子宫内膜癌的AUC为0.967,显著大于ADC值的0.898(P=0.0336)。
- 预测复发的AUC随机森林模型0.890 vs ADC值0.875,差异无统计学意义(P=0.7248)。
- ADC值≤0.802×10⁻³ mm²/sec组无复发生存期显著短于>0.802×10⁻³ mm²/sec组(P<.0001;HR 32.9)。
核心结论
- ADC图纹理分析预测组织学分级的诊断性能显著高于ADC值,且对无复发生存期的预测比组织学分级和ADC值更有用。
研究设计与样本
- 单中心研究,纳入121例子宫内膜癌患者,采用1.5-T MRI系统及b值为0和1000 sec/mm²的扩散加权成像。
- 在ADC图上以感兴趣体积提取45个纹理特征(含高阶纹理特征),采用受试者工作特征分析比较随机森林模型与ADC值预测组织学分级和复发的诊断性能。
诊断标准与关键数值
- 预测高级别子宫内膜癌的AUC:随机森林模型0.967 vs ADC值0.898(P=0.0336)。
- 预测复发的AUC:随机森林模型0.890 vs ADC值0.875(P=0.7248),差异无统计学意义。
- ADC值阈值0.802×10⁻³ mm²/sec用于无复发生存期分析。
临床应用与阅片要点
- ADC图纹理分析(随机森林模型)可用于预测子宫内膜癌组织学分级,其诊断性能优于单纯ADC值。
- ADC值≤0.802×10⁻³ mm²/sec与更短的无复发生存期相关(P<.0001;HR 32.9)。
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: To determine the feasibility of texture analysis (TA) of apparent diffusion coefficient (ADC) maps for predicting histologic grade (HG) and recurrence-free survival (RFS) in patients with endometrial carcinoma (EMC).
MATERIALS AND METHODS: One hundred twenty-one patients with EMC were examined by using a 1.5-T MRI system and diffusion-weighted imaging (DWI) with b values of 0 and 1000 sec/mm2. Software with volumes of interest on ADC maps was used to extract 45 texture features including higher-order texture features. Receiver operating characteristic analysis was performed to compare the diagnostic performance of the random forest (RF) model and ADC values for HG and recurrence.
RESULTS: Area under the curve (AUC) for predicting high-grade EMCs was significantly larger for RF model than for ADC values (0.967 vs 0.898; P = .0336). AUC for predicting recurrence was larger for the RF model than for ADC values (0.890 vs 0.875; P = .7248), although the difference was not significant. Mean RFS was significantly shorter for high-grade EMCs than for low-grade EMCs (P = .0002; hazard ratio, 4.9) and for ADC values less than or equal to 0.802 × 10-3 mm2/sec than for ADC values greater than 0.802 × 10-3 mm2/sec (P < .0001; hazard ratio, 32.9). RF model showed that the mean RFS was significantly shorter for the presence of recurrence than for its absence (P < .0001; hazard ratio, 94.7).
CONCLUSION: TA of ADC maps had significantly higher diagnostic performance than did ADC values for predicting HG and was a more useful indicator than HG and ADC values for predicting RFS in patients with EMC.Keywords: Comparative Studies, Genital/Reproductive, MR-Diffusion Weighted Imaging, MR-Imaging, Neoplasms-Primary, Pathology, Pelvis, Tissue Characterization, Uterus© RSNA, 2019.
引用
- Yamada I, Miyasaka N, Kobayashi D, et al. Endometrial Carcinoma: Texture Analysis of Apparent Diffusion Coefficient Maps and Its Correlation with Histopathologic Findings and Prognosis. Radiol Imaging Cancer. 2019. PMID: 33778684. doi:10.1148/rycan.2019190054
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。