面向髋部骨折风险预测的高灵敏度与广泛适用性的两阶段模型
机器学习
2025-10-20 v1 医学物理
摘要
髋部骨折是老年人群体中导致残疾、死亡以及医疗负担的主要原因,因此凸显了早期风险评估的必要性。然而,常用工具如DXA T-score和FRAX常常缺乏灵敏度,难以识别无既往骨折或骨质疏松者的高风险人群。为此,本文提出一种序列式两阶段模型,整合临床与影像信息以提高预测准确性。以Osteoporotic Fractures in Men Study (MrOS)、Study of Osteoporotic Fractures (SOF)以及UK Biobank为数据来源,阶段1(筛查)采用临床、人口学及功能变量估算基线风险,而阶段2(影像)则融合DXA提取的特征进行精细化。该模型通过内部与外部测试进行严格验证,在不同队列中表现出稳定且具适应性。与T-score和FRAX相比,两阶段框架实现了更高的灵敏度并减少了漏诊病例,为早期髋部骨折风险评估提供了具有成本效益且个性化的方案。
引用
@article{arxiv.2510.15179,
title = {An Advanced Two-Stage Model with High Sensitivity and Generalizability for Prediction of Hip Fracture Risk Using Multiple Datasets},
author = {Shuo Sun and Meiling Zhou and Chen Zhao and Joyce H. Keyak and Nancy E. Lane and Jeffrey D. Deng and Kuan-Jui Su and Hui Shen and Hong-Wen Deng and Kui Zhang and Weihua Zhou},
journal= {arXiv preprint arXiv:2510.15179},
year = {2025}
}
备注
38 pages, 3 figures, 8 tables. This is a preprint version of the manuscript titled "An Advanced Two-Stage Model with High Sensitivity and Generalizability for Prediction of Hip Fracture Risk Using Multiple Datasets." The paper is currently under journal submission