Objectives To construct a prediction model related to MRI total burden score of cerebral small vessel disease （CSVD）， so as to realize the use of individual indicators to predict the severity of CSVD in imaging. Methods This study is a retrospective study， including 429 CSVD patients in outpatient department and inpatient department of Neurology in Affiliated Hospital of Xuzhou Medical University from July 2018 to March 2019. General clinical data （including demographic characteristics such as gender， age， vascular risk factors such as hypertension history， diabetes history， coronary heart disease history， stroke history， smoking and drinking， etc.）， laboratory examination （including blood routine， liver and kidney function， blood lipid， coagulation function， inflammatory index， thyroid function， etc.） and imaging data （such as brain MRI） of all patients were collected. According to the total CSVD score， the patients were divided into mild group （score 0 to 1， 198 cases）， moderate group （score 2， 93 cases） and severe group （score 3 to 4， 138 cases）.Single factor analysis was used to compare the data differences between the groups. With the mild group as a reference， multivariate Logistic regression was used to analyze the factors influencing the occurrence of CSVD in the moderate or the severe group， and ROC was used to analyze the value of each factor in predicting the moderate or severe CSVD. Results Compared with mild CSVD，multivariate Logistic regression showed that age （OR=1.115，95%CI：1.077-1.156），history of hypertension （OR=2.549，95%CI：1.393-4.662）， AST （OR=0.953， 95%CI；0.911-0.997） and cystatin C （OR=53.246， 95%CI：2.774-1021.986） were predictors of moderate CSVD （P＜ 0.05）， while age （OR=1.156，95%CI：1.113-1.200）， history of hypertension （OR=4.642，95%CI：2.425- 8.883），creatinine （OR=1.029，95%CI：1.007-1.052）， HDL （OR=0.312，95%CI：0.102-0.952）， lymphocyte count （OR=0.4406，95%CI：0.243-0.797），APTT （OR=1.158，95%CI：1.046-1.282） and HCY （OR=1.119， 95%CI：1.025-1.220） were predictors of severe CSVD （P＜ 0.05）. By plotting the ROC curve of the prediction probability obtained from the above model and the selected single variable，the AUCROC of joint diagnosis of moderate CSVD is 0.828（95%CI：0.777-0.878）， and the AUCROC of combined diagnosis of severe CSVD is 0.910 （95%CI：0.880-0.940）.The AUCROC of combined diagnosis of moderate/severe CSVD was larger than that of single factor AUCROC，and the difference was statistically significant （P＜ 0.05）. Conclusions Understanding the demographic information， past history，laboratory tests and other information of CSVD patients to build a prediction model of CSVD severity is more conducive to the prediction of moderate/severe CSVD than a single index