基于LASSO回归和SSA-BP神经网络模型对川崎病冠状动脉病变的预测研究

打开文本图片集

关键词：SSA-BP神经网络模型；LASSO回归；川崎病；冠状动脉病变 中图分类号：R725.4;TP18 文献标识码：A DOI：10.3969/j.issn.1006-1959.2025.24.001

文章编号：1006-1959（2025）24-0001-09

Prediction of Coronary ArteryLesions in Kawasaki Disease Based onLASSO Regression and SSA-BP Neural Network Model

LEI Yi，LI Baofei， WANG Xinfang， CHEN Zhaoyang （Departmentofediatrics，ZhongshanHospitalofTraditional ChineseMedicine/theTenth Clinical MedicalColegeofGuangzhou University of Chinese Medicine，Zhongshan 5284O0，Guangdong，China）

Abstract：Obeaeosfstei），ep neuranetworkaocht（Suralkdletlalpe MethodsClincaldatafro53Kawasakidsesehdren（CALositie，4CALegativetreatedatZogshanHospitalofTadiiole MedicinebetweenMarch21andMay25wereanaled.KeypredictorserescreeedusingLASOregresso.WithCALcomplcatioaste dependenarabledtasesereslittoagnstigsat8n：tiraletkodelosrucedyB wascompareditheraditioalacheaingodelThefectivessofeodelaerfdyteestsetdataditsfoce （acuracyesiifiUC.）aedulsaueleieedaie （IVIG）espoeeactieotesebliaatelsodatlaeleotO） P= （20 0.008）and erythrocyte sedimentation rate（ P=0.005 were not selected as variables.When the training set：test set=8：2，the SSA-BP model showed the best performance on the training set，with an accuracy rate of 95.24% and an F1-score of 0.8889，which were the highest among all models.On the test set，although the accuracy of SSA-BP，LR and BP models was 81.82% ，the overall performance of SSA-BP model was more balanced.The KNN andSVMmodelsereompleelyuabletodentifALpositiecassothteststesitivityaO）;hentetraingsestte advantage of SSA-BP model was more obvious.On the test set，the accuracy of the SSA-BP model was 93.75% ，theaccuracy and specificity were （2 100% ，the sensitivity was 80% ，and theF1-score was O.6667.Conclusion SSA-BP model shows stable and excellent prediction performance under 7：3 paritionicoatSifiesbsofoatdsutllif traditional BP neural network.However，there is still a 20% risk of CAL identification in clinical practice.In the future，it is necessary to integrate keyfactorssunddcdstteesilallble diagnostic svstem.

Key Words：SSA-BP neural network model; LASSO regression;Kawasaki disease;Coronary artery lesions

川崎病（Kawasakidisease，KD）好发于5岁以下儿童，是一种不明原因的、急性自限性血管炎性疾病，主要累及中型动脉，尤其是冠状动脉病变（coronaryarterylesion，CAL）可导致冠状动脉扩张、冠状动脉瘤以及发生冠状动脉血栓形成或狭窄甚至猝死等恶性冠状动脉事件I。（剩余13942字）