抑郁症疗效的深度学习预测模型  

Deep learning in predicting the treatment outcomes of depressed patients

作　　者：陈晓瑜[1] 胡逸儒 张宾 Chen Xiaoyu;Hu Yiru;Zhang Bin(Psychological and Psychiatric Neuroimage Lab,Affiliated Brain Hospital of Guangzhou Medical University,Guangzhou 510370,China)

机构地区：[1]广州医科大学附属脑科医院精神心理脑功能实验室,广州510370

出　　处：《中华行为医学与脑科学杂志》2022年第11期1041-1045,共5页Chinese Journal of Behavioral Medicine and Brain Science

基　　金：广东省科技计划项目(2019B030316001)。

摘　　要：不同抑郁症患者的最优治疗方案一般需通过长期、低效率的试错过程来逐步确定。为实现抑郁症的精准治疗,有必要通过特异性生物标志物来选择有效的治疗方法。深度学习是机器学习的一个分支,该技术能处理大量高维、复杂的数据,适用于自动提取和学习临床、基因组学和神经影像数据的特征。近年来,研究人员正在使用深度学习技术开发抑郁症治疗反应的预测模型,有利于指导临床医生为患者选择最佳治疗方案以及在全球范围内推进更为高效的个体化精准医疗方案。本文从人口学、临床症状数据、基因组学数据和功能磁共振成像数据三个方面,对深度学习预测抑郁症疗效方面的相关研究进行综述,并对未来的深度学习研究方向尤其是多组学数据结合深度学习的应用进行展望。The optimal antidepressant therapies for different patients have been identified mostly by trial and error.Selecting an effective treatment based on the specific biomarkers may be an important step toward personalized treatment of depression.Deep learning is a branch of machine learning,that is capable of processing high-dimensional and complex data.It automatically extracts and learns from large amounts of demographic,clinical symptoms,genomics and neuroimaging data.Researchers have been using deep learning algorithms to develop prediction model of anti-depressant response in order to guide clinicians to make a precise prescription for depression and further advance personalized treatment globally.This article reviews the application of deep learning in predicting the treatment outcomes of depression.Additionally,deep learning based on multi-omics data applied in treatment outcome's prediction is direction with prospects in the future.

关 键 词：深度学习 预测模型 抑郁症 基因组学 功能磁共振成像 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] R749.4[自动化与计算机技术—控制科学与工程]