【作者】 高源；

【导师】 李劲松；

【作者基本信息】 浙江大学 ， 生物医学工程， 2019， 硕士

【摘要】 结直肠肿瘤是发病率与死亡率都较高的肿瘤之一。在结直肠肿瘤的诊治过程中,临床预后预测发挥着重要的作用。医生可以根据特定的临床指南或预测模型,来预测患者接下来可能发生的特定预后事件。目前预后的预测主要基于专家的临床经验和共识,缺乏循证支持。随着近年来电子健康记录的大规模应用和计算机领域内机器学习算法的不断发展,基于大规模临床数据的预测模型逐渐在预后预测中发挥越来越重要的作用。本论文基于机器学习中的经典算法逻辑回归模型,借鉴在医学领域内广泛应用的多因素Meta回归分析的思想,以及以贝叶斯先验概率理论为基础的模型聚合框架,构建了适合多种情况、应用于多种场景的结直肠肿瘤预后模型,并使用公开数据集与真实数据集相结合的方式,设计两组实验从不同的角度对这些模型进行验证。在此基础上,以上述模型为核心,多种Web应用开发技术为工具,设计并实现了结直肠肿瘤预后预测系统。该系统可以根据训练数据的来源不同,由研究者或者医生团队自定义设置研究方案和模型训练方案,以满足用户的研究及预后预测等多种需求。系统的服务端集成了数据存储模块与数据分析模块,客户端与服务端则以Web应用服务的方式进行数据交换,可以方便地为处于不同地域的团队提供支持。更多还原

【Abstract】 Colorectal cancer(CRC)is one of the cancers with high morbidity and mortality.Clinical prognosis plays an important role in the diagnosis and treatment of colorectal cancer.Physicians can predict specific prognostic events that may occur next in the patient based on specific clinical guidelines or predictive models.The current prognosis is mainly based on the expert’s clinical experience and consensus,and lacks evidence-based support.With the large-scale application of electronic health records in recent years and the continuous development of machine learning algorithms in the computer field,predictive models based on large-scale clinical data are playing an increasingly important role in prognosis prediction.This thesis is based on the classical algorithm logistic regression model in machine learning,inspired from the multi-factor meta-regression analysis,which is widely used in the medical field,and the model aggregation framework based on Bayesian prior probability theory,which is suitable for a variety of situations.In a variety of scenarios of colorectal cancer prognosis models,and using a combination of public data sets and real data sets,two sets of experiments were designed to validate these models from different perspectives.On this basis,with the above model as the core,a variety of network application development techniques as tools,this thesis designed and implemented a colorectal cancer prognosis prediction system.The system can be customized by the researcher or doctor team to set up research plans and model training programs according to the source of the training data to meet the needs of users’ research and prognosis prediction.The server side of the system integrates the data storage module and the data analysis module,and the client and the server exchange data in the manner of network application service,which can conveniently support the teams in different regions.更多还原