机器学习筛选生物标志物诊治下肢深静脉血栓的研究进展

Machine learning predicts biomarkers for assistant diagnosis and treatment of deep venous thrombosis of lower limbs

下肢深静脉血栓是常见的外周血管疾病,其发病率有上升趋势,并可导致肺栓塞、脑梗塞等严重并发症,甚至危及生命。目前临床早期预防、诊断下肢深静脉血栓主要依赖于影像学检查,存在局限性。通过机器学习筛选生物标志物如D-二聚体、纤溶酶原激活物抑制物-1、Stat3、凝血酶原时间、内皮素-1、纤维蛋白原、凝血因子Ⅴ和Ⅷ及血浆高密度脂蛋白胆固醇、P选择素、血管细胞黏附蛋白-1、基质金属蛋白酶-8和肝细胞生长因子等用于辅助诊断下肢深静脉血栓,以及使用机器学习算法建立模型选择及评价治疗方案和药物成为了一种新思路,具有准确率高、敏感性高、检查费用相对较低的特点,可为临床医生提供参考。

Lower extremity deep vein thrombosis(DVT)is a common peripheral vascular disease and its incidence has an upward trend. DVT can lead to pulmonary embolism,cerebral infarction and other serious and life-threatening complications. At present,early clinical prevention and diagnosis of lower limb deep vein thrombosis mainly rely on imaging examination,which has limitations. Through machine learning,biomarkers such as D-dimer,plasminogen activator inhibitor-1,Stat3,prothrombin time,endothelin-1,fibrinogen,coagulation factor Ⅴ and Ⅷ,plasma high-density lipoprotein cholesterol,P-selectin,vascular cell adhesion protein-1,matrix metalloproteinase-8,and hepatocyte growth factor were screened for auxiliary diagnosis of lower limb deep vein thrombosis. Using machine learning algorithm to establish models to select and evaluate treatment plans and drugs has become a new idea,which has the characteristics of high accuracy,high sensitivity and relatively low inspection cost,and can provide reference for clinicians.