基于半监督nnU-Net的[82Rb]PET/CT左心室全自动分割
研究在40例非门控[82Rb]PET/CT系列(20例患者,静息与负荷)上建立整合动态PET与CT的手动多模态分割协议作为参考标准,并纳入805例额外无标注动态PET系列(504例患者)进行半监督学习。nnU-Net平均Dice达87.8%[85.6,89.2],显著优于优化半自动阈值法基线75.1%[72.9,76.9]%,在低灌注区域召回率提升20–30%。结论认为深度学习可实现接近专家水平的全自动左心室分割,支持大规模心肌血流量(MBF)定量。
要点
- nnU-Net平均Dice 87.8%,显著优于阈值法基线75.1%。
- 低灌注区域召回率较阈值法提升20–30%。
- 半监督学习利用805例无标注系列 modestly 提升静息与负荷鲁棒性。
核心结论
- 深度学习可实现[82Rb]PET/CT左心室全自动分割,接近专家精度,支持大规模MBF定量。
研究设计与样本
- 单中心40例非门控[82Rb]PET/CT系列(20例患者,静息与负荷)建立手动多模态分割参考标准。
- 采用五折交叉验证,并纳入805例额外无标注动态PET系列(504例患者)进行半监督学习。
- 与优化半自动阈值法基线(35%)比较模型性能。
主要发现
- nnU-Net平均Dice 87.8%[85.6,89.2],基线75.1%[72.9,76.9]。
- nnU-Net召回率89.1%[86.1,91.4],基线82.6%[79.1,85.4]。
- nnU-Net精确率88.1%[84.2,90.4],基线70.2%[67.2,73.0]。
- 低灌注区域召回率较阈值法提升20–30%。
- 半监督学习在静息与负荷采集上均适度提升模型鲁棒性。
关键数值与推荐
- 参考标准队列包含广泛坏死 challenging 病例,中位坏死比例21%。
- 半自动阈值法基线采用35%阈值。
临床怎么用
- 可消除手动交互,用于大规模MBF定量与可重复的高通量心脏PET分析。
局限性
- 参考标准仅基于单中心40例系列,样本量有限。
- 半监督学习仅适度提升鲁棒性。
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)。
原文
BACKGROUND: Quantification of myocardial blood flow (MBF) with [Formula: see text]Rb PET/CT requires accurate delineation of the left ventricle (LV). Manual or semi-automated contouring remains time-consuming and error-prone, particularly in hypoperfused myocardium. We developed and validated a fully automatic LV segmentation pipeline using nnU-Net applied to [Formula: see text]Rb PET/CT. A manual, multimodal segmentation protocol integrating dynamic PET and CT was established in a single-center cohort of 40 non-gated PET/CT series (20 patients, rest & stress), including challenging cases with extensive necrosis (median 21%). The resulting ground truth masks were used for five-fold cross-validation, and semi-supervised learning incorporated 805 additional unlabeled dynamic PET series (504 patients). Model performance was compared with an optimized semi-automatic thresholding baseline (35% [Formula: see text]). RESULTS: The nnU-Net significantly outperformed the baseline, achieving a mean Dice of 87.8[85.6, 89.2]% vs 75.1[72.9, 76.9]%, recall 89.1[86.1, 91.4]% vs 82.6[79.1, 85.4]%, and precision 88.1[84.2, 90.4]% vs 70.2[67.2, 73.0]%. The improvement was most pronounced in hypoperfused regions, where recall increased by 20-30% compared to thresholding. Semi-supervised learning modestly enhanced model robustness across both rest and stress acquisitions. CONCLUSIONS: A deep-learning-based approach enables fully automatic LV segmentation in [Formula: see text]Rb PET/CT with near-expert accuracy. This framework eliminates manual interaction, supports large-scale MBF quantification, and paves the way for reproducible, high-throughput cardiac PET analysis in clinical and research workflows.
引用
- Amirian M, Chevalley A, Asiain MM, et al. Fully automatic left ventricle segmentation in [Formula: see text]Rb PET/CT Using a semi-supervised nnU-net. EJNMMI Res. 2026. PMID: 42207374. doi:10.1186/s13550-026-01455-3
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。