基于先验医学知识的风险预测模型

A Risk Prediction Model Based on Prior Medical Knowledge

通过电子健康记录预测潜在疾病风险任务是近年来医疗领域的研究热点。随着机器学习研究与应用的快速发展,经典机器学习模型渐渐无法满足日益增长的数据量和复杂的数据分析需求,而深度学习中神经网络模型可以解决机器学习无法解决或难以解决的问题。现有疾病预测工作中没有对先验医学知识的明确考虑。本文提出了一种新的、通用的框架,称为风险预测任务,它可以使用后验正则化技术成功地将离散的先验医学知识应用到所有最先进的预测模型中。本文以卷积神经网络建立风险预测模型,并加入先验医学知识,以梯度下降算法进行优化。实验证明,与传统深度学习中的卷积神经网络相比该模型能有效提高风险预测的准确率。

The task of predicting potential disease risks through electronic health records is a hot research topic in the medical field in recent years. With the rapid development of machine learning research and application, the classical machine learning model is gradually unable to meet the growing data volume and complex data analysis needs, while the neural network model in deep learning can solve the problem that machine learning cannot or is difficult to solve. There is no explicit consider-ation of prior medical knowledge in existing disease prediction work. We propose a new, generic framework called the risk prediction task, which successfully applies discrete prior medical knowledge to all the most advanced prediction models using posterior regularization techniques. In this paper, the convolution neural network is used to establish the risk prediction model, and prior medical knowledge is added. Gradient descent algorithm was used for optimization. Experiments prove that this model can effectively improve the accuracy of risk prediction compared with the convolution neural network in traditional deep learning.