基于机器学习的神经精神疾病辅助诊断研究进展-《中国医学物理学杂志》

文章信息/Info

Title:: Advances in auxiliary diagnosis of neuropsychiatric diseases based on machine learning

作者:: 雷炳业1; 2; 潘嘉瑜2; 吴逢春2; 3; 陆小兵2; 3; 宁玉萍2; 3; 陈军4; 5; 吴凯1; 2; 3; 4; 5; 1.华南理工大学材料科学与工程学院生物医学工程系，广东广州 510006； 2.广东省精神疾病转化医学工程技术研究中心，广东广州 510370； 3.广州医科大学附属脑科医院（广州市惠爱医院），广东广州 510370； 4.广东省老年痴呆诊断与康复工程技术研究中心，广东广州 510500； 5.国家医疗保健器具工程技术研究中心，广东广州 510500

Author(s):: LEI Bingye1; 2; PAN Jiayu2; WU Fengchun2; 3; LU Xiaobing2; 3; NING Yuping2; 3; CHEN Jun4; 5; WU Kai1; 2; 3; 4; 5; 1. Department of Biomedical Engineering, School of Material Science and Engineering, South China University of Technology, Guangzhou 510006, China; 2. Guangdong Engineering Technology Research Center for Translational Medicine of Mental Disorders, Guangzhou 510370, China; 3. the Affiliated Brain Hospital of Guangzhou Medical University (Guangzhou Huiai Hospital), Guangzhou 510370, China; 4. Guangdong Engineering Technology Research Center for Diagnosis and Rehabilitation of Dementia, Guangzhou 510500, China; 5. National Engineering Technology Research Center for Healthcare Devices, Guangzhou 510500, China

Keywords:: Keywords: neuropsychiatric disease; neuroimaging; machine learning; auxiliary diagnosis

摘要:: 【摘要】神经影像技术被广泛应用于研究大脑结构和功能异常与神经精神疾病之间的相关性。与传统的统计学分析方法不同，机器学习模型能对神经影像学数据进行个体化预测，发掘潜在的生物学标记物。神经精神疾病辅助诊断包含数据预处理和机器学习算法。数据预处理是一种人为的特征工程，为机器学习算法提供量化特征；机器学习算法包含特征降维、模型训练和模型评估。鲁棒的机器学习算法可以实现对不同数据集的准确预测，并提供对预测结果贡献大的特征，作为潜在的生物学标记物。本文综述了近年来基于机器学习的神经精神疾病辅助诊断研究进展，从数据预处理、机器学习算法和生物学标记物3个角度进行介绍，并展望未来的研究方向。【关键词】神经精神疾病；神经影像；机器学习；辅助诊断

Abstract:: Abstract: Neuroimaging techniques are widely used to study the correlations between brain structural/functional abnormalities and neuropsychiatric diseases. Different from traditional statistical analysis methods, machine learning model can realize individualized prediction from neuroimaging data and exploit potential biomarkers. The auxiliary diagnosis of neuropsychiatric diseases includes data preprocessing and machine learning algorithms. Data preprocessing is a kind of artificial feature engineering, providing quantitative features for machine learning algorithms; and machine learning algorithms include feature dimensionality reduction, model training and model evaluation. Robust machine learning algorithms can accomplish accurate predictions for different datasets and provide features that contribute significantly to the prediction as potential biomarkers. Herein the recent advances in auxiliary diagnosis of neuropsychiatric diseases based on machine learning are summarized, including data preprocessing, machine learning algorithms and biomarkers found in previous studies. Finally, the research direction in the future is discussed.