基于医疗大数据的阿尔茨海默病症状发展预测模型

Understanding and modeling cognitive status of Alzheimer’s patients or identifying those at risk of developing Alzheimer’s disease during the pre-symptomatic phase is pivotal in our attempts to assist real world clinical decision-making and clinical trial designs. Currently, brain imaging data are acquired routinely in hospitals providing access to large datasets which currently exist (as well as future data) available without requiring additional resources or expertise. One major route to model the mental status is to use brain-imaging data and significant efforts have been devoted in this area. However, a sophisticated model that accurately predicts the mental status of individuals is not yet available.. Preliminary researches of the applicant also indicate that clinical data and genetic data are effective when measure the disease development. Thus, combing clinical data, genetic data and brain-imaging data is very promising. The proposed research will focus on two aspects: (1) Using computational models to predict the Mini Mental State Examination (MMSE) and other cognitive scores with brain image data, clinical data and genetic data. The result will be also used to measure cognitive declines on those at risk of developing Alzheimer’s disease. (2) Establish computational models to classify participants into three diagnostic groups with high level of accuracy: Alzheimer’s Disease, Mild Cognitive Impairment, Normal.. The applicant has significant experience theoretically and experimentally in the interdisciplinary fields of computer science and its applications, and bioinformatics, which is going to lead to the state-of-the art achievements.

借助机器学习算法帮助医生理解阿尔兹海默症的发病原因，发现潜在疾病的高危人群，预测患者病情的发展趋势，对疾病的诊断和治疗具有重要意义和实用价值。目前，关于阿尔兹海默症病辅助诊断的研究方法主要依赖于医院设备容易获取的人脑图像数据，但至今尚无成熟模型可以有效预测患者的认知状态变化。申请者前期研究证明：临床医学数据和遗传学数据也能有效地评估和预测疾病的发展。本项目研究涵盖2个内容。其一，提出融合多源异构的人脑图像数据、临床医学数据和遗传学数据的多种特征，构建具有原创性的计算生物学预测模型，通过该模型预测患者的核心认知指标MMSE以及认知分数的变化，预测潜在人群患病概率等。其二，依托医学大数据构建辅助诊断分类模型，通过该模型将被诊断人群分为三类：阿尔兹海默症患者、轻度认知障碍者和正常人。申请人在计算机应用和生物医学交叉领域积累了扎实的基础理论和丰富的科研经验，有望获得高水平的原创性研究成果。

借助机器学习算法帮助医生理解阿尔兹海默症的发病原因，预测患者病情的发展趋势，对疾病的诊断和治疗具有重要意义和实用价值。目前，关于阿尔兹海默症病辅助诊断的研究方法主要依赖于医院设备容易获取的人脑图像数据，但至今尚无成熟模型可以有效预测患者的认知状态变化。. 项目研究涵盖2个内容。其一，研究融合人脑图像数据、临床医学数据和遗传学数据的多种特征，构建计算生物学预测模型，预测患者的核心认知指标MMSE以及认知分数的变化等。其二，研究基于医学数据的辅助诊断分类模型，通过模型将被诊断人群分为三类：阿尔兹海默症患者、轻度认知障碍者和正常人。. 项目研究取得的原创性成果包括：基于孪生脑磁共振成像的阿尔兹海默症诊断模型，基于排序卷积神经网络的阿尔兹海默症患者病情发展预测模型，基于相似性的阿尔兹海默症诊断模型，基于遗传因子的风湿性关节炎患者药物效果预测模型，基于自然邻居和自标记半监督学习的不平衡分类算法及其在聚类分析、异常检测和实例选择中的应用。相关成果的呈现形式包括：. ①学术论文：课题组成员在Computer Methods and Programs in Biomedicine (IF=5.428, JCR-1区), Arthritis & Rheumatology (IF=10.995, JCR-1区), Information Sciences (IF=6.795, JCR-1区), Knowledge-Based Systems (IF=8.038, JCR-1区), Applied Intelligence (IF=5.086, JCR-1区), Egyptian Informations Journal (IF=3.943)，Experimental Cell Research (IF=3.905) 等国际期刊上发表反映项目研究成果、并标注受到本项目资助的SCI期刊论文10篇。其中，影响因子IF>5的JCR-1区高水平论文7篇。. ②发明专利：结合本项目研究所取得的进展和成果，课题组成员申请国际和国内发明专利7项。其中，WIPO国际发明专利3项，国家发明专利4项。