文献速递眼整形与眼眶神经眼科

深度学习眼球突出测量系统在非甲状腺眼眶病中的泛化评估

Generalisation of a deep learning-based exophthalmometry system using facial photographs for non-thyroidal orbital disease.
PubMed 文献 · Br J Ophthalmol · 2026-09-15

本研究回顾性纳入269人(95名健康对照、174例单侧非甲状腺眼眶病患者),评估仅用甲状腺眼病队列训练的Glandy EXO深度学习系统基于面部照片的自动眼球突出测量与手动Hertel测量的的一致性。结果显示两组一致性良好,系统对眼间不对称(>2.0 mm)的敏感度为59.6%、特异度95.0%、准确度83.3%。

要点

  • TED训练的自动测量系统在非TED眼眶病及健康眼中与Hertel测量一致性良好。
  • 系统检测眼间不对称(>2.0 mm)的敏感度59.6%、特异度95.0%、准确度83.3%。
  • 84.8%的自动测量值在手动Hertel值2.0 mm范围内。

核心结论

  • 仅用TED队列训练的深度学习眼球突出测量系统在健康人和非TED眼眶病患者中与Hertel测量一致性良好,可作为临床检查的辅助工具用于眼球位置的客观量化和记录。

研究设计与样本

  • 回顾性观察性研究,共纳入269人,包括95名健康对照和174例单侧非甲状腺眼眶病患者。
  • 自动测量使用Glandy EXO系统基于同日正面面部照片获得,与手动Hertel眼球突出测量进行比较。
  • 一致性评估采用Pearson相关系数(PCC)、平均绝对误差(MAE)、组内相关系数(ICC)和Bland-Altman分析,并评估对临床显著眼间不对称(>2.0 mm)的诊断性能。

主要发现

  • 健康对照眼:PCC 0.80,MAE 0.86 mm,ICC 0.80。
  • 患者患眼:PCC 0.81,MAE 1.41 mm,ICC 0.75。
  • 患者健眼:PCC 0.76,MAE 1.03 mm,ICC 0.75。
  • 总体84.8%的测量值在手动Hertel值2.0 mm范围内。
  • 检测临床显著眼间不对称(>2.0 mm)的敏感度59.6%、特异度95.0%、准确度83.3%。

关键数值与推荐

  • 自动测量与Hertel测量的一致性以PCC、MAE、ICC和Bland-Altman分析评估。
  • 临床显著眼间不对称定义为>2.0 mm。
  • 健康对照眼PCC 0.80、MAE 0.86 mm、ICC 0.80;患眼PCC 0.81、MAE 1.41 mm、ICC 0.75;健眼PCC 0.76、MAE 1.03 mm、ICC 0.75。
  • 总体84.8%的测量值在手动Hertel值2.0 mm范围内。
  • 检测眼间不对称(>2.0 mm)的敏感度59.6%、特异度95.0%、准确度83.3%。

临床怎么用

  • 该系统可作为临床检查的辅助工具,用于眼球位置的客观量化和记录。
  • 在非甲状腺眼眶病患者中,自动测量与Hertel测量一致性良好,但检测眼间不对称的敏感度有限(59.6%)。
  • 基于面部照片的自动测量无需额外训练即可在非TED眼眶病中应用。

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

AIMS: To evaluate agreement between photo-based automated exophthalmometry and manual Hertel exophthalmometry in healthy individuals and patients with non-thyroidal orbital disease, with a deep learning system trained exclusively in thyroid eye disease (TED) cohorts applied without retraining or fine-tuning. METHODS: This was a retrospective observational study. A total of 269 individuals were included in the study, including 95 healthy controls and 174 patients with unilateral non-TED orbital pathology. Automated measurements obtained using Glandy EXO from same-day frontal facial photographs were compared with corresponding manual Hertel exophthalmometry measurements. Agreement was assessed using Pearson correlation coefficient (PCC), mean absolute error (MAE), intraclass correlation coefficient (ICC) and Bland-Altman analysis. Diagnostic performance for clinically significant intereye asymmetry (>2.0 mm) was also evaluated. RESULTS: The system showed good agreement with Hertel exophthalmometry in the eyes of healthy controls (PCC 0.80; MAE 0.86 mm; ICC 0.80) as well as the affected (PCC 0.81; MAE 1.41 mm; ICC 0.75) and fellow (PCC 0.76; MAE 1.03 mm; ICC 0.75) eyes of patients. Overall, 84.8% of measurements were within 2.0 mm of the manual Hertel values. The sensitivity, specificity and accuracy of the system for detecting clinically significant intereye asymmetry were 59.6%, 95.0% and 83.3%, respectively. CONCLUSIONS: The TED-trained photoexophthalmometry system showed good agreement with Hertel measurements in healthy individuals and patients with non-TED orbital disease. These findings support its use as an adjunct to clinical examination for objective quantification and documentation of globe position.

来源:Br J Ophthalmol,查看原文。版权归原作者所有。

引用

  1. Park J, Byeon HJ, Shin K, et al. Generalisation of a deep learning-based exophthalmometry system using facial photographs for non-thyroidal orbital disease. Br J Ophthalmol. 2026. PMID: 42744591. doi:10.1136/bjo-2026-330182

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。