欢迎访问药物流行病学杂志网站！

设为首页加入收藏

国家科技部中国科技论文统计源期刊

中国科技核心期刊

WHO西太平洋地区医学索引(WPRIM)收录期刊

湖北优秀期刊

《药物流行病学杂志》官方网站与投审稿系统变更通知<点击查看详情>

人工神经网络模型预测药物性肝损伤的临床转归及影响因素MIV值评价

Artificial Neural Network Prediction Model Predict the Clinical Prognosis of Drug induced Liver Injury And Evaluate the Importance of MIV Values for the Relevant Influencing Factors

DOI：

中文关键词: 人工神经网络药物性肝损伤转归预测模型平均影响值

英文关键词:Artificial neural network Drug induced liver injury Prognosis prediction model Mean impact value

基金项目:

作者	单位
张琦¹ 李青² 冷光² 赵伟²	1.山西医科大学药学院太原030001；2.山西医科大学附属第一医院药学部

摘要点击次数: 1333

全文下载次数: 569

中文摘要:

摘要目的：利用人工神经网络方法，对药物性肝损伤（DILI）的转归建立人工神经网络反向传播（Back propagation Artificial Nerual Network，BP ANN）预测模型，预测DILI患者的临床转归并对相关影响因素进行平均影响值（mean impact value，MIV）评价。方法：收集某三甲医院全部科室2014年6月1日～2017年6月1日以“药物性肝损害”、“药物性肝损伤”、“药物性肝炎”、“药物性肝病”、“药物性肝衰竭”、“药物性肝硬化”其中之一为主要诊断的266例住院患者资料。符合纳排标准的，对其临床各项指标与转归情况之间的关联性进行Spearman相关性分析，筛选出具有相关性的指标将其作为输入神经元，将转归情况作为输出神经元，构建BP ANN模型。模型构建训练完毕后，从2017年7月1日起收集70例符合纳入排除标准的DILI住院患者进行临床转归预测，并监测其实际转归情况，将预测结果与实际结果进行对比。进一步用已经训练好的BP ANN预测模型对有相关性的指标即影响因素进行MIV评价，分析各影响因素对于DILI影响的重要性大小。结果：266例住院患者中，最终符合纳入排除标准的有190例。Spearman相关性分析结果显示共有17项指标有统计学意义，提示有相关性。预测结果显示70例患者中有64例的预测转归与实际转归相符，模型预测的符合率为91.43%。经BP ANN分析，根据MIV值，直接胆红素首次异常值、血清白蛋白含量、γ 谷氨酰转肽酶首次异常值、体重指数、天冬氨酸转氨酶首次异常值是对DILI患者临床转归影响最大的5个相关指标。结论：人工神经网络模型预测药物性肝损伤临床转归符合率较高，药物性肝损伤的临床转归大部分趋于痊愈或好转。

英文摘要:

ABSTRACT Objective:The artificial neural network method was used to establish the back propagation artificial neural network prediction model for predicting the clinical prognosis of drug induced liver injury（DILI）, and evaluating the importance of MIV values for the relevant influencing factors. Methods:266 hospitalized patients whose main diagnosis was “drug induced liver damage” or “drug induced liver injury” or “drug induced hepatitis” or “drug induced liver disease” or “drug induced liver failure” or “drug induced cirrhosis” were selected from all departments in a top three hospital in June 1, 2014 June 1, 2017. According to the inclusion exclusion criteria, the correlation between clinical indicators and prognosis was analyzed by Spearman correlation. We screened out the relevant index as input neurons, the prognosis as output neurons, and the artificial neural network back propagation (BP ANN) model was constructed. After the completion of the model construction training, from July 1, 2017, 70 inpatients with DILI who meet the inclusion criteria were collected to predict the clinical outcome. We monitored their actual outcome, then the forecast results and the actual results were compared. Furthermore, MIV evaluation was made on the related indexes and the influencing factors by the trained BP neural network model, and the importance of each influencing factor on the DILI was analyzed. Results:Of the 266 hospitalized patients, 190 eventually met the inclusion criteria. The results of the Spearman correlation analysis showed that there were 17 indicators with statistical significance, suggesting a correlation. The prediction results showed that the prediction outcome of 64 out of 70 patients was consistent with the actual outcome, and the coincidence rate of the model prediction was 91.43%. By BP ANN analysis, according to MIV value, the first abnormal value of direct bilirubin, serum albumin, the first abnormal value of γ glutamyl transferase, body mass index and the first abnormal value of aspartate aminotransferase were 5 relevant indicator which the most impact on clinical outcomes of patients with DILI. 〖WTHZ〗Conclusion:〖WTBZ〗The artificial neural network model had a high coincidence rate in predicting the of drug induced liver injury. Most of the clinical outcomes of drug induced liver injury tend to be cured or improved.

查看全文查看/发表评论下载PDF阅读器

关闭