Han Jingyang, Wang Tao, Du Xiaoyan, Wang Yali, Guo Ziyang, Li Dandan, Yu Xinjun
Geriatric Medicine Department, Affiliated Hospital of Shandong Second Medical University, Weifang, China.
School of Clinical Medicine, Affiliated Hospital of Shandong Second Medical University, Weifang, China.
Front Neurol. 2025 Aug 7;16:1636696. doi: 10.3389/fneur.2025.1636696. eCollection 2025.
Benign paroxysmal positional vertigo (BPPV) is the most common type of vertigo in clinical practice. Previous studies have suggested that inflammatory responses and metabolic disorders may be involved in the pathogenesis of BPPV, but systematic analyses based on large samples are lacking. The aim of this study is to construct an intelligent auxiliary diagnostic model for BPPV based on the big data of SRM-IV vertigo diagnostic and treatment system, and to carry out clinical validation.
The clinical data of 522 vertigo patients were retrospectively analyzed, including 303 BPPV patients and 219 non-BPPV patients. LASSO regression and random forest algorithm were used to screen feature variables, and based on the screened feature variables, multifactor logistic regression analysis was performed to establish a prediction model for BPPV auxiliary diagnosis. Finally, the model was applied to BPPV patients diagnosed by SRM-IV diagnosis and treatment system for external validation.
Multifactorial logistic regression analysis showed that disease duration, neutrophils, lymphocytes, C-reactive protein (CRP), ferritin, and vitamin D deficiency were independent risk factors for the diagnosis of BPPV (OR>1, < 0.05), monocyte count was an independent protective factor for the diagnosis of BPPV (OR<1, < 0.05), and the area under curve (AUC) was 0.927.
The intelligent assisted diagnostic model of BPPV constructed based on the big data of SRM-IV vertigo diagnostic and treatment system has high diagnostic accuracy and clinical application value, and it is expected to assist the clinicians to improve the diagnostic efficiency.
良性阵发性位置性眩晕（BPPV）是临床实践中最常见的眩晕类型。以往研究提示炎症反应和代谢紊乱可能参与BPPV的发病机制，但缺乏基于大样本的系统分析。本研究旨在基于SRM-IV眩晕诊疗系统的大数据构建BPPV智能辅助诊断模型，并进行临床验证。
回顾性分析522例眩晕患者的临床资料，其中BPPV患者303例，非BPPV患者219例。采用LASSO回归和随机森林算法筛选特征变量，并基于筛选出的特征变量进行多因素logistic回归分析，建立BPPV辅助诊断预测模型。最后，将该模型应用于SRM-IV诊疗系统诊断的BPPV患者进行外部验证。
多因素logistic回归分析显示，病程、中性粒细胞、淋巴细胞、C反应蛋白（CRP）、铁蛋白及维生素D缺乏是BPPV诊断的独立危险因素（OR>1，<0.05），单核细胞计数是BPPV诊断的独立保护因素（OR<1，<0.05），曲线下面积（AUC）为0.927。
基于SRM-IV眩晕诊疗系统大数据构建的BPPV智能辅助诊断模型具有较高的诊断准确性和临床应用价值，有望协助临床医生提高诊断效率。
