卷积神经网络模型在儿科疾病预测中的应用

Application of Convolutional Neural Network in Pediatric Disease Prediction

目的:针对儿童看病需求量大导致的儿科诊疗服务效率和准确率偏低等问题,利用自然语言处理和深度学习技术,从儿科历史病历数据中自动"学习"专家医生诊断模式,形成智能辅助诊断模型,从而对新的儿科病历数据输出疾病诊断决策。结果:基于深度卷积神经网络的七分类疾病智能诊断模型的正确率为84.26%,F1-score为84.33%,基本达到可投入实际应用的级别。结论:智能诊断决策作为预诊信息提供给医生进行确诊参考,对提升医生诊断速度效果明显。

Objective： To solve the inferior efficiency and accuracy of pediatric diagmosis because of the large demand for children＇s medical treatnlent, by natural language processing and deep learning technology to automatically ＂learn＂ the expert doctor diagnosis mode based on pediatric historical medical record data, and then establishes an intelligent auxiliary diagmosis model. The model can provide disease diagnosis decisions for new pediatric medical record. Results： The accuracy of intelligent diagnosis model based on deep convolutional neural network with respect to seven-classification application is 84.26% and F 1-score is 84.33%, which initially achieves the practical level. Conclusion： The intelligent diagnosis decision serves as pre-diagnosis and can offer to the ＆）ctor for reference, which can greatly improve the speed of clinical diagnosis.