Based on Machine Learning Algorithm: Construction of an Early Prediction Model of Integrated Traditional Chinese and Western Medicine for Cognitive Impairment after Ischemic Stroke

摘要

Purpose: Based on the risk factors of post stroke cognitive impairment (PSCI), combining the Constitution and Syndrome of Traditional Chinese Medicine, using a variety of Machine learning (ML) algorithms, to construct a prediction model with high accuracy and good fitting degree, so as to provide theoretical and data support for early screening and early prevention of ischemic stroke (IS) patients. Patients and methods: A retrospective analysis was conducted on 85 patients with acute ischemic stroke admitted to the Department of Neurology of a third grade a hospital of integrated Traditional Chinese and Western Medicine (TCM-WM) from June 2019 to January 2020. The patients were divided into three groups: Support Vector Machine (SVM), Random Forest (RF), Gradient Boosting Decision Tree (GBDT), ML algorithms were used to construct the risk prediction model of post-stroke cognitive impairment, and the prediction accuracy and area under curve (AUC) of receiver operating characteristic curve (ROC) were used to evaluate the prediction effect of the three models. Results: The average prediction accuracy of GBDT was 80.77 percent, the highest and the most stable. The average AUC area of GBDT was 0.85, which was larger than that of the other three ML algorithms, and the prediction effect was better. After analyzing the importance of the features obtained from the training of GBDT model, it is concluded that the features with the highest degree of discrimination for PSCI in this data set are as follows: Barthel index, Age, fasting blood glucose (FPG), blood homocysteine (Hcy). Based on GBDT algorithm, four GBDT models were obtained by training 75 percent, 80 percent, 85 percent and 90 percent training sets respectively. It was found that the prediction accuracy of the models with 85 percent and 90 percent training sets could reach 84.62 percent and 88.89 percent, indicating the potential of applying machine learning algorithm to the prediction of cognitive impairment after ischemic stroke. Conclusion: The ML algorithm is used to construct the early prediction model of TCM-WM integration for cognitive impairment after ischemic stroke, and analyze the influencing factors with strong correlation with PSCI, so as to carry out early detection, early diagnosis and early treatment of PSCI, so as to provide basis and reference for researchers who construct a large sample prediction model of cognitive impairment after ischemic stroke.