文献速递腹部CT

nn-UNet TotalSegmentator 肝脏分割在非典型肝脏中的性能分析

Performance analysis of liver segmentation using nn-UNet TotalSegmentator: Focus on atypical livers, pathologies, and variants.
PubMed 文献 · Eur J Radiol · 2026-01-14

本研究回顾性纳入 67 例健康肝脏 CT 与 55 例涵盖 11 类病理/变异亚组的肝脏 CT,评估 nn-UNet TotalSegmentator 自动分割准确性。以手动分割并经两名放射科医师审核结果为参考,采用 Dice 相似系数、Hausdorff 距离、平均表面距离、容积差及临床评分评价。结果显示该工具在健康肝脏表现优异,在多数病理肝脏表现良好,但多囊肝等亚组临床可用性受限。

要点

  • 自动分割低估健康肝脏容积平均 48.11 ml(3.1%),高估 84% 病理/变异肝脏容积平均 79.09 ml(4%)。
  • 多囊肝、肝硬化伴腹水、獭尾肝 Dice 评分显著低于健康肝组,其中多囊肝最低(0.656±0.230)。
  • Dice 评分高并不等于临床可用,獭尾肝、肝硬化伴腹水、消融缺损、转移、半肝切除亚组临床评分更低。

核心结论

  • TotalSegmentator 在健康肝脏上表现优异,在多数病理肝脏上表现良好,但多囊肝、獭尾肝等亚组存在明显局限。
  • Dice 评分在很多病理病例中较高,但临床评分揭示其实际可用性受限,临床评估仍不可或缺。
  • 在训练数据中纳入多囊肝与獭尾肝病例可能减少偏倚并改善分割性能。

研究设计与样本

  • 回顾性收集 RIS 中的 CT 扫描,分为参考组(67 例健康肝脏)与研究组(55 例,覆盖 11 个病理与变异亚组)。
  • 以手动分割并经两名放射科医师审核校正的结果作为参考标准,评估 nn-UNet TotalSegmentator 自动分割表现。
  • 评价指标包括 Dice 相似系数、Hausdorff 距离、平均表面距离、容积差及临床评分。

关键影像表现/征象

  • 自动分割在 84% 的研究组病例中高估肝脏容积,平均高估 79.09 ml(4%)。
  • 参考组自动分割平均低估肝脏容积 48.11 ml(3.1%)。

诊断标准与关键数值

  • 参考组平均 Dice 评分为 0.980±0.007,研究组平均 Dice 评分为 0.933±0.113。
  • 肝大亚组 Dice 评分最高(0.979±0.006),多囊肝亚组 Dice 评分最低(0.656±0.230)。
  • 与参考组相比,肝硬化伴腹水、獭尾肝及多囊肝的 Dice 评分显著更低。

临床应用与阅片要点

  • 对健康肝脏及多数常见病理肝脏可采用 TotalSegmentator 辅助容积评估,但报告前仍需逐层审阅分割边界。
  • 多囊肝、獭尾肝、肝硬化伴腹水、消融缺损、转移及半肝切除病例不宜直接采信自动分割结果。
  • Dice 评分较高的病理病例仍可能出现临床不可接受的分割错误,应结合临床评分判断可用性。

局限性

  • 研究为回顾性单中心设计,样本量有限,尤其是各病理亚组病例数较少。

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

RATIONALE AND OBJECTIVES: This study evaluates the accuracy of the nn-UNet TotalSegmentator (TS) by Wasserthal et al. (2023) in segmenting atypical livers with pathologies and variants in CT scans.

MATERIALS AND METHODS: CT scans were retrospectively collected from our RIS and divided into two cohorts: a reference group (67 healthy livers) and a study group (55 scans across eleven pathology and variant subgroups). TS performed automatic segmentation for all groups. For reference, the images were then manually segmented, with corrections reviewed by two radiologists. Accuracy was assessed using Dice similarity score, Hausdorff distance (HD), mean surface distance (MSD), volume difference, and clinical ratings.

RESULTS: Automatic segmentation underestimated liver volume by a mean of 48.11 ml (3.1%) in the reference group and overestimated it in 84% of study group cases by 79.09 ml (4%). The average Dice score was 0.980 ± 0.007 for the reference group and 0.933 ± 0.113 for the study group. Hepatomegaly achieved the highest score (0.979 ± 0.006), Polycystic liver disease (PLD) the lowest (0.656 ± 0.230). Cirrhosis with Ascites, Beavertail, and PLD had significantly lower Dice scores than the reference group. Clinical ratings were often lower than Dice scores suggested, especially in Beavertail, Cirrhosis with Ascites, Ablation defects, Metastases, and Hemihepatectomy.

CONCLUSION: TS performs excellently on healthy and well on most pathological livers. Despite high Dice scores in many pathological cases, clinical ratings reveal limitations. Clinical evaluation remains essential. Inclusion of PLD and Beavertail cases in training data may reduce bias and improve performance.

来源:Eur J Radiol,查看原文。版权归原作者所有。

引用

  1. Kleiß, Arndt, Sommerfeld, et al. Performance analysis of liver segmentation using nn-UNet TotalSegmentator: Focus on atypical livers, pathologies, and variants. Eur J Radiol. 2026. PMID: 41581492. doi:10.1016/j.ejrad.2026.112674

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。