人工神经网络在预测输尿管结石自行排出中的应用研究

发布时间：2018-06-05 04:03

本文选题：输尿管结石 + 神经网络(计算机)　；参考：《中国全科医学》2016年20期

【摘要】：目的运用人工神经网络建立输尿管结石自行排出的预测模型,并转化成临床应用。方法选取2013年1—8月在石河子大学医学院第一附属医院泌尿外科门诊就诊的输尿管结石患者225例。经保守排石治疗4周后,复查泌尿系B超或CT判断结石是否排出,并将患者分为排石组和未排石组。通过单因素分析筛选出影响结石排出的因素,将这些因素作为预测参数建立人工神经网络,并对68例测试集样本进行预测。绘制预测拟概率的ROC曲线,并计算ROC曲线下面积评价预测效能。结果排石组141例,未排石组84例。单因素分析结果显示两组患者性别、体质指数、膀胱刺激征、侧别、肾盂积水、尿p H值、血尿、淋巴细胞计数比较,差异均无统计学意义(P0.05);两组患者年龄、疼痛程度、结石直径、结石位置、血白细胞计数、中性粒细胞计数、中性粒细胞分数、淋巴细胞分数、C反应蛋白水平比较,差异均有统计学意义(P0.05)。运行人工神经网络,输入层共建立9个神经元。系统自动体系构建两个隐含层,输出层有1个神经元。预测变量重要性前3位分别是:结石直径(0.20)、C反应蛋白(0.18)、年龄(0.12)。运用建立成功的人工神经网络对68例测试集样本进行预测,结果显示,测试集样本的灵敏度、特异度和总准确率分别为93.3%、60.9%和82.4%,ROC曲线下面积为0.868〔95%CI(0.774,0.962)〕。结论人工神经网络预测输尿管结石能否排出有较高的准确率,可辅助临床医师为患者制定安全、合理的治疗方案。
[Abstract]:Objective to establish a prediction model of ureteral calculi excretion by artificial neural network (Ann) and transform it into clinical application. Methods 225 patients with ureteral calculi were selected from January to August 2013 in Urology Department of the first affiliated Hospital of Shihezi University Medical College. After 4 weeks of conservative lithotomy treatment, the patients were divided into two groups: lithotomy group and non-lithotomy group. Factors affecting stone excretion were screened by single factor analysis. Artificial neural network was established by using these factors as predictive parameters and 68 test samples were predicted. The ROC curve of the prediction quasi probability is drawn and the area under the ROC curve is calculated to evaluate the prediction efficiency. Results there were 141 cases in the lithotomy group and 84 cases in the non-lithotomy group. Univariate analysis showed that there were no significant differences in sex, BMI, bladder irritation, side, hydronephrosis, urine pH, hematuria and lymphocyte count between the two groups. There were significant differences in stone diameter, stone location, leukocyte count, neutrophil fraction, lymphocyte fraction and C-reactive protein level (P 0.05). An artificial neural network was run and nine neurons were established in the input layer. The system automatically constructs two hidden layers with one neuron in the output layer. The first three positions of predictive variable importance were: stone diameter 0.20 C reactive protein 0.18, age 0.12. A successful artificial neural network was used to predict 68 test samples. The results showed that the sensitivity, specificity and total accuracy of the test samples were 93.30.9% and 82.4%, respectively. The area under the ROC curve was 0.86895%. Conclusion artificial neural network can predict whether ureteral calculi can be excreted with high accuracy and can assist clinicians to make safe and reasonable treatment plan for patients with ureteral calculi.
【作者单位】：石河子大学医学院第一附属医院泌尿外科;
【基金】：国家科技支撑计划子课题(2013BAI05B05) 新疆生产建设兵团医药卫生重点项目科技援疆专项(2014AB052)
【分类号】：R693.4

【相似文献】