医疗大数据驱动的疾病预测系统架构与应用研究

Research on the Architecture and Applications of Medical Big Data-Driven Disease Prediction Systems

摘要: 本文针对医学影像识别与临床风险预测两类典型场景，设计并构建了一套基于医疗大数据的疾病预测原型系统，给出了由数据处理层、AI模型层、应用服务层和用户交互层组成的四层系统架构。在影像端，系统以预训练的ResNet50为骨干网络，通过迁移学习完成胸部X光片的肺炎识别，并使用Softmax输出类别概率；在临床风险预测方面，系统集成逻辑回归、随机森林和XGBoost等机器学习模型，实现冠心病风险概率及风险概率预测。同时引入SHAP方法对模型决策进行解释，增强了结果的可信度与可用性。

Abstract: This paper focuses on two typical scenarios—medical image recognition and clinical risk prediction—and designs and implements a disease prediction prototype system based on healthcare big data. A four-layer system architecture is proposed, consisting of a data processing layer, an AI model layer, an application service layer, and a user interaction layer. For the imaging component, the system employs a pre-trained ResNet50 as the backbone network, performs pneumonia recognition on chest X-ray images through transfer learning, and outputs class probabilities using Softmax. For clinical risk prediction, the system integrates machine learning models including Logistic Regression, Random Forest, and XGBoost to predict coronary heart disease risk and associated probability. Additionally, the SHAP method is introduced to interpret model decisions, thereby enhancing the credibility and usability of the results.