Method for predicting relapse of cerebral arterial thrombosis

Abstract

The invention provides a method for predicting relapse of cerebral arterial thrombosis, which comprises the following steps of: firstly, multi-dimensional data of a patient are extracted for fusion, Lasso analysis is performed on the fused data, and key factors are output; secondly, vacancy values in the data set are filled, and three different modes are adopted for filling of patients which do not relapse and have hospitalization histories, features which have large missing quantities of the patients which have no hospitalization histories and features which have small missing quantities of the patients which have no hospitalization histories; the sample imbalance existing in the data set is processed by adopting a sample imbalance processing mode; meanwhile, CT image data of the brain of a patient are taken, convolutional learning is conducted on the image data through a GCForest multi-granularity scanning layer, and features are normalized to be in the size of [32, 1] through feature remodeling; the remodeled features are taken as fixed features and structured features, the fixed features and the structured features are jointly transmitted to a GCForest multi-granularity scanning layer for feature enhancement, and the features are finally transmitted to a cascade forest for model training. A new thought is provided for application of the artificial intelligence technology in medical treatment.

Description

technical field [0001] The invention relates to the technical field of ischemic stroke recurrence prediction, in particular to a method for ischemic stroke recurrence prediction. Background technique [0002] Stroke is the leading cause of death and disability in my country, with a recurrence rate as high as 14.7%. The risk of death and disability after recurrence of ischemic stroke is 9.4 times that of the first stroke. Recurrence risk prediction helps to identify high-risk groups for stroke recurrence, and provides decision-making information support for three-early prevention. There are three main types of current medical prediction technology: one is traditional machine learning algorithms, such as: Logistic regression, SVM, decision tree, etc.; the other is deep neural network, such as: multi-layer perceptron MLP, LSTM, GRU, etc.; Another category is integrated algorithms, such as: random forest, Adboost, Xgboost, etc. Appeal Medical prediction methods and techniques ...

AI Technical Summary

Problems solved by technology

But there is still no general method that can be applied to solve the medical prediction problem

Moreover, there are relatively few studies on the prediction of recurrence after discharge of ischemic stroke patients. In this research field, researchers have not conducted research on data filling of ischemic stroke data gaps and sample imbalance. Second, researchers have not Fusion of imaging data and inpatient structured feature data to build a model

Images