盆腔超声提肌裂孔分割的多UNet静态加权集成模型
本研究回顾性纳入186例盆腔器官脱垂女性,行经会阴超声并构建提肌裂孔分割模型。训练、验证、测试集分别为106、37、43例,以Dice相似系数与Jaccard指数评价。最佳集成模型MoUNetsUNet&ACCUNet的DSC为0.9653、JI为0.9333,较多数单一模型显著提升,但与标准UNet无统计学差异。
要点
- 盆腔超声提肌裂孔自动分割中,标准UNet在单一模型中一致性最好。
- MoUNets集成对较弱模型提升更明显,适合提升整体分割稳健性。
- 最佳集成模型与标准UNet无显著差异,临床选用需权衡计算成本。
核心结论
- 静态加权集成多个UNet专家模型可提升较弱模型对提肌裂孔的分割性能,在困难病例中有明确临床价值。
- 最佳集成模型MoUNetsUNet&ACCUNet的DSC为0.9653、JI为0.9333。
- 最佳集成模型与标准UNet比较无统计学显著差异,p=0.888。
- 最佳集成模型与其余五个单一模型比较均有统计学显著差异,均p<0.01。
研究设计与样本
- 186例患者中位年龄33.5岁,四分位间距30.0至48.8岁。
- 数据随机分为训练集106例、验证集37例、测试集43例。
- 采用Dice相似系数和Jaccard指数在测试集上评价模型分割性能,并在不同架构间比较。
诊断标准与关键数值
- 单一模型中标准UNet的提肌裂孔分割一致性最好。
- 最佳集成模型MoUNetsUNet&ACCUNet的DSC为0.9653。
- 最佳集成模型MoUNetsUNet&ACCUNet的JI为0.9333。
临床应用与阅片要点
- 盆腔超声提肌裂孔分割可优先考虑标准UNet作为基线模型。
- 若临床部署中已有多个性能相近的分割专家模型,可采用静态加权集成提升较弱模型的稳定性。
- 在困难病例中集成策略比单一较弱模型更具优势,但需注意其对强模型提升有限。
局限性
- 本研究为单中心回顾性设计,样本量为186例,泛化能力有待验证。
- 最佳集成模型与标准UNet在统计上无显著差异,实际临床增量有限。
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: This study aimed to investigate whether a static weighted ensemble of multiple UNet-based expert models (MoUNets) can achieve more accurate segmentation of the levator hiatus in pelvic floor ultrasound images than any single expert model.
METHODS: This retrospective single-center study included women diagnosed with pelvic organ prolapse who underwent transperineal ultrasound examinations between July 2020 and November 2023. The cohort was randomly divided into the training (n = 106), validation (n = 37) and testing (n = 43) cohorts. Model performance was evaluated on the test set using the Dice similarity coefficient (DSC) and Jaccard index (JI) and compared across all competing architectures.
RESULTS: The study included 186 women (median age, 33.5 y [IQR, 30.0-48.8 y]). Comparative experiments on our in-house levator hiatus dataset confirmed that MoUNets is effective. Among the six conventional UNet models, the standard UNet achieved the best segmentation consistency. MoUNets, which combines different expert models of similar performance, showed higher numerical values than individual expert models. The best ensemble, MoUNetsUNet&ACCUNet, yielded a DSC of 0.9653 and a JI of 0.9333. However, paired t-tests revealed no statistically significant difference between the ensemble and the UNet (p = 0.888). In contrast, statistically significant differences were observed between the best ensemble model and each of the other five individual models (all p < 0.01).
CONCLUSION: The ensemble model leverages complementary advantages among different experts to significantly improve the segmentation performance of weaker models, particularly in difficult cases, demonstrating clear clinical value.
引用
- Zhang, Wu, Lou, et al. MoUNets: Static-Weighted Ensemble of UNet-Based Experts for Levator Hiatus Segmentation in Pelvic Floor Ultrasound. Ultrasound Med Biol. 2026. PMID: 42731935. doi:10.1016/j.ultrasmedbio.2026.08.018
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。