Research And Implementation Of Computer-Aided Decision-Making In Diagnosis And Treatment Of Intracerebral Hemorrhage

Intracerebral hemorrhage is a serious cerebrovascular disease,which often leads to death or disability of patients,and imposes heavy burdens to families and society.Analysis of risk factors for intracerebral hemorrhage,recommendation of treatment options,and prediction of patient outcomes play an important role in the process of preventive diagnosis,patient treatment,and disease assessment.To address the problems of complex patient etiology and numerous risk factors in brain hemorrhage risk factor analysis,multi-level dependence of patient consultation information in treatment plan recommendation,and complex modalities of modeling data in patient outcome prediction,this thesis conducts computer-aided decision-making research on intracerebral hemorrhage diagnosis and treatment based on machine learning and artificial intelligence technology,and proposes new solutions to the above problems in order to improve the efficiency of diagnosis and treatment of brain hemorrhage patients.The main work of the thesis is as follows:1.Aiming at the problem that intracerebral hemorrhage involves complex risk factors and the search algorithm is trapped in local optimum,a two-stage gray wolf sensing optimization algorithm(TGWSOA)is proposed,in which the first stage uses the idea of hierarchy to stratify the attributes by domain knowledge,and each stratum max-relevance and min-redundancy(m RMR)algorithm is used to obtain the important risk factors in each stratum,and then the candidate set of risk factors is obtained by merging.The second stage uses the grey wolf optimization algorithm with perception and variation to perform secondary screening of the set to determine the final set of risk factors.The experimental results showed that the classification accuracy of the twostage gray wolf perceptual optimization algorithm reached 91.25% and 70.42% for the cardiovascular disease dataset and the brain hemorrhage dataset,respectively.2.Aiming at the problem of lack of dependence between visits in the treatment plan recommendation model,a hierarchical feedback interaction network model(Hierarchical Feedback Interaction Networks,HIFINet)is proposed.The model is divided into inspection layer,diagnosis layer and treatment layer.Between layers A differential feedback unit is added,and the simulated doctor diagnoses and prescribes treatment drugs for patients according to the test results.The experimental results show that on the MIMIC-III dataset,the JACCARD score of HIFINet reaches 0.5215,and the F1 score reaches 0.6755.3.Aiming at the problems of complex patient data modality and insufficient data interaction in the patient death risk prediction model,a central core memory network model(CCMN)is proposed,which captures the temporal correlation within the sequence by embedding,and adopts the central core memory module Capturing correlations between multimodal time-varying series.The experimental results show that on the MIMIC-III dataset and the matching subset of the MIMIC-III waveform database,the AUC-ROC of CCMN reaches 0.9095,and the AUC-PRC reaches 0.3155.4.Based on frameworks such as Spring Boot and Vue,using My SQL database,and applying B/S architecture design to realize the decision support platform for cerebral hemorrhage diagnosis and treatment,the system has functions such as patient electronic medical record management,risk factor analysis,treatment plan recommendation and death risk prediction,and the operation effect is good.It can better assist doctors in decision-making.