传统影像组学与VGG-SAM融合策略对宫颈癌MRI术前神经周围侵犯预测的比较评估
本研究回顾性纳入103例宫颈癌患者（82例PNI阴性、21例PNI阳性），比较传统影像组学与VGG-SAM深度学习策略基于MRI术前预测神经周围侵犯（PNI）的性能。最强影像组学基线（PyRadiomics+XGBoost）的ROC-AUC为0.6731，而朴素VGG+SAM融合达到最佳敏感度0.5714、平衡准确度0.7491、ROC-AUC 0.7854；可学习融合特异度最高（0.9634）。混合VGG-SAM建模优于传统影像组学，但融合复杂度增加未带来一致获益。
要点
- 朴素VGG+SAM融合在敏感度和ROC-AUC上优于传统影像组学基线，适合筛查导向。
- 可学习融合将操作点移向高特异度（0.9634），适合排除导向。
- 融合复杂度增加未带来一致获益，需多中心大样本验证。
核心结论
- 混合VGG-SAM建模优于单独传统影像组学，但更高融合复杂度未产生一致获益。
- 朴素融合偏向敏感度导向判别，可学习融合主要将操作点移向更高特异度。
- 需要更大规模多机构队列和临床对照验证。
研究设计与样本
- 回顾性队列研究，纳入103例宫颈癌患者，其中82例PNI阴性、21例PNI阳性。
- 从原始DICOM MRI及配对肿瘤掩膜提取PyRadiomics特征。
- 评估四种深度变体：仅VGG、仅SAM、VGG+SAM朴素融合、VGG+SAM可学习融合。
- 采用患者水平分层五折交叉验证。
- 主要指标为敏感度、特异度、平衡准确度、F1分数、ROC-AUC和PR-AUC；不确定性用10,000次配对bootstrap重采样和精确McNemar检验评估。
诊断标准与关键数值
- 最强影像组学基线PyRadiomics+XGBoost：平衡准确度0.6057，F1分数0.3684，ROC-AUC 0.6731，PR-AUC 0.4460。
- 朴素VGG+SAM融合：敏感度0.5714，平衡准确度0.7491，F1分数0.6154，ROC-AUC 0.7854，PR-AUC 0.6231，为最佳。
- 可学习融合：特异度0.9634，精确率0.7500，准确度0.8544（与最高准确度持平）。
- 配对bootstrap比较显示临床相关指标置信区间宽。
临床应用与阅片要点
- 术前MRI预测PNI可辅助神经保留手术决策，但需注意本研究样本量小、PNI阳性病例少。
- 若追求高敏感度筛查PNI，可参考朴素VGG+SAM融合策略；若追求高特异度排除，可参考可学习融合策略。
- 传统影像组学单独预测性能有限，混合深度学习模型可能提供补充价值。
局限性
- 回顾性单中心研究，样本量有限（103例），PNI阳性仅21例。
- 配对bootstrap比较置信区间宽，结果稳定性不足。
- 缺乏多机构外部验证和临床对照。
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: Perineural invasion (PNI) is an adverse feature in cervical cancer and may influence nerve-sparing surgery. We compared conventional radiomics and VGG-SAM deep-learning strategies for MRI-based preoperative prediction of PNI.
METHODS: A retrospective cohort of 103 patients with cervical cancer, including 82 PNI-negative and 21 PNI-positive patients, was analyzed. PyRadiomics features were extracted from raw DICOM MRI and paired tumor masks. Four deep variants were evaluated: VGG only, SAM only, VGG + SAM naive fusion, and VGG + SAM learnable fusion. Patient-level stratified five-fold cross-validation was used. Primary metrics were sensitivity, specificity, balanced accuracy, F1-score, ROC-AUC, and PR-AUC. Uncertainty for the naive-versus-learnable comparison was assessed with 10,000 paired bootstrap resamples and exact McNemar tests.
RESULTS: The strongest radiomics baseline, PyRadiomics + XGBoost, achieved balanced accuracy 0.6057, F1-score 0.3684, ROC-AUC 0.6731, and PR-AUC 0.4460. Naive VGG + SAM fusion achieved the best sensitivity (0.5714), balanced accuracy (0.7491), F1-score (0.6154), ROC-AUC (0.7854), and PR-AUC (0.6231). Learnable fusion achieved the highest specificity (0.9634) and precision (0.7500) while matching the highest accuracy (0.8544). Paired bootstrap comparisons showed wide confidence intervals across clinically relevant metrics.
CONCLUSION: Hybrid VGG-SAM modeling outperformed conventional radiomics alone, but greater fusion complexity did not yield uniform benefit. Naive fusion favored sensitivity-oriented discrimination, whereas learnable fusion mainly shifted the operating point toward higher specificity. Larger multi-institutional cohorts and clinical comparators are needed.
引用
- Tan L, Xia Y, Teng D, et al. Comparative evaluation of conventional radiomics and VGG-SAM fusion strategies for MRI-based preoperative prediction of perineural invasion in cervical cancer. Abdom Radiol (NY). 2026. PMID: 42171740. doi:10.1007/s00261-026-05559-1
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。