用常规实验室检测数据判定肾脏病理的神经网络方法  

Application of Artificial Neural Networks in Predicting Renal Histopathology

作　　者：沈延春[1] 朱忠华[2] 杨晓[2] 庹新兰[1] 喻盼[1] 黄蓉[1] 王劲[1] 

机构地区：[1]江汉大学医学院,湖北武汉430056 [2]华中科技大学同济医学院附属协和医院肾内科,湖北武汉430022

出　　处：《临床医学工程》2010年第12期26-29,共4页Clinical Medicine & Engineering

摘　　要：目的通过临床上常规的血液和尿液的检查结果,运用人工神经网络(Artificial Neural Networks,ANN)预测肾脏病理情况。方法收集1998年～2005年有血常规、肝肾功能、血脂、免疫球蛋白补体、出凝血功能及24h尿蛋白检查并有肾穿刺病理诊断的92例,应用神经网络对肾脏反应增生、浸润和硬化程度的25项指标的预测。结果预测值和实际值之间无显著差异(P>0.05)。结论人工神经网络对肾脏疾病中的肾脏病理情况的预测具有可行性和实用价值。Objective To explore the prospect of predicting the renal histopathology to use the result of blood and urine by artificial neural network（ANN） . Methods 92 cases collected from 1998 to 2005 had the results of blood routine,liver and renal function,serum lipids,immunoglobulin and complement,blot and cruor function,24-hour urine protein and the histopathology diagnosis of renal biopsy. To predict 25 indexs about hyperplasia,infiltrate and sclerosis by ANN. Results The value of prediction and truth had no marked difference（P 0.05） . Conclusion ANN can predict the renal histopathology in kidney disease. The analytic method of ANN has advantage and availability in the relation of renal clinicopathological domain.

关 键 词：人工神经网络 肾脏病理 实验室检查 

分 类 号：R692[医药卫生—泌尿科学]