Palm Viktoria, Thangamani Subasini, Budai Bettina Katalin, Skornitzke Stephan, Eckl Kira, Tong Elizabeth, Sedaghat Sam, Heußel Claus Peter, von Stackelberg Oyunbileg, Engelhardt Sandy, Kopytova Taisiya, Norajitra Tobias, Maier-Hein Klaus H, Kauczor Hans-Ulrich, Wielpütz Mark Oliver
Clinic for Diagnostic and Interventional Radiology (DIR), Heidelberg University Hospital, Heidelberg, Germany.
Diagnostic and Interventional Radiology with Nuclear Medicine, Thoraxklinik Heidelberg, Heidelberg, Germany.
Sci Rep. 2025 Jul 1;15(1):20756. doi: 10.1038/s41598-025-05091-0.
Predicting vertebral height is complex due to individual factors. AI-based medical imaging analysis offers new opportunities for vertebral assessment. Thereby, these novel methods may contribute to sex-adapted nomograms and vertebral height prediction models, aiding in diagnosing spinal conditions like compression fractures and supporting individualized, sex-specific medicine. In this study an AI-based CT-imaging spine analysis of 262 subjects (mean age 32.36 years, range 20-54 years) was conducted, including a total of 3117 vertebrae, to assess sex-associated anatomical variations. Automated segmentations provided anterior, central, and posterior vertebral heights. Regression analysis with a cubic spline linear mixed-effects model was adapted to age, sex, and spinal segments. Measurement reliability was confirmed by two readers with an intraclass correlation coefficient (ICC) of 0.94-0.98. Female vertebral heights were consistently smaller than males (p < 0.05). The largest differences were found in the upper thoracic spine (T1-T6), with mean differences of 7.9-9.0%. Specifically, T1 and T2 showed differences of 8.6% and 9.0%, respectively. The strongest height increase between consecutive vertebrae was observed from T9 to L1 (mean slope of 1.46; 6.63% for females and 1.53; 6.48% for males). This study highlights significant sex-based differences in vertebral heights, resulting in sex-adapted nomograms that can enhance diagnostic accuracy and support individualized patient assessments.
由于个体因素，预测椎体高度很复杂。基于人工智能的医学影像分析为椎体评估提供了新机会。因此，这些新方法可能有助于制定针对性别的列线图和椎体高度预测模型，有助于诊断诸如压缩性骨折等脊柱疾病，并支持个体化的、针对性别的医学。在本研究中，对262名受试者（平均年龄32.36岁，范围20 - 54岁）进行了基于人工智能的CT影像脊柱分析，共包括3117个椎体，以评估性别相关的解剖学变异。自动分割提供了椎体的前部、中部和后部高度。采用三次样条线性混合效应模型进行回归分析，以适应年龄、性别和脊柱节段。两名阅片者确认测量可靠性，组内相关系数（ICC）为0.94 - 0.98。女性椎体高度始终小于男性（p < 0.05）。在上胸椎（T1 - T6）发现最大差异，平均差异为7.9 - 9.0%。具体而言，T1和T2的差异分别为8.6%和9.0%。从T9到L1观察到相邻椎体之间最强的高度增加（女性平均斜率为1.46；6.63%，男性平均斜率为1.53；6.48%）。本研究强调了椎体高度存在显著的性别差异，从而产生了针对性别的列线图，可提高诊断准确性并支持个体化的患者评估。
