当前位置: X-MOL 学术Cognit. Comput. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Functional Brain Network Classification for Alzheimer’s Disease Detection with Deep Features and Extreme Learning Machine
Cognitive Computation ( IF 5.4 ) Pub Date : 2019-11-29 , DOI: 10.1007/s12559-019-09688-2
Xin Bi , Xiangguo Zhao , Hong Huang , Deyang Chen , Yuliang Ma

The human brain can be inherently modeled as a brain network, where nodes denote billions of neurons and edges denote massive connections between neurons. Analysis on functional brain networks provides powerful abilities to discover potential mechanisms of human brain, and to aid brain disease detection, such as AD (Alzheimer’s disease). Effective discrimination of patients of AD and MCI (mild cognitive impairment) from NC (normal control) is important for the early diagnosis of AD. Therefore, this paper explores the problem of brain network classification for AD detection. Two deep learning methods of functional brain network classification are designed. The convolutional learning method learns the deep regional-connectivity features, while the recurrent learning method learns the deep adjacent positional features. The ELM (extreme learning machine)-boosted structure is also implemented to further improve the learning ability. Extensive experiments are conducted to evaluate and compare the AUC (area under curve), accuracy, recall, and training time of the proposed methods on a real-world dataset. Results indicate that (1) the proposed methods which learn deep features directly from brain networks outperform shallow learning methods and (2) models with the ELM-boosted structure achieve a higher performance. This paper explores the brain networks learning with deep features and ELM. The results demonstrate that the proposed methods provide a satisfactory learning ability in the application of AD detection.

中文翻译:

具有深度特征和极限学习机的阿尔茨海默氏病检测功能脑网络分类

可以将人类大脑固有地建模为大脑网络,其中节点表示数十亿个神经元,边缘表示神经元之间的大量连接。对功能性大脑网络的分析提供了强大的功能,可以发现人脑的潜在机制,并有助于诸如AD(阿尔茨海默氏病)之类的脑部疾病检测。有效区分AD和MCI(轻度认知障碍)与NC(正常对照)的患者对于AD的早期诊断很重要。因此,本文探讨了用于AD检测的脑网络分类问题。设计了两种功能脑网络分类的深度学习方法。卷积学习方法学习深度的区域连通性特征,而递归学习方法学习深度的邻近位置特征。还实施了ELM(极限学习机)增强结构,以进一步提高学习能力。进行了广泛的实验,以评估和比较实际数据集上所提出方法的AUC(曲线下面积),准确性,召回率和训练时间。结果表明:(1)提出的直接从脑网络学习深度特征的方法优于浅层学习方法;(2)具有ELM增强结构的模型实现了更高的性能。本文探索具有深层功能和ELM的脑网络学习。结果表明,所提出的方法在AD检测的应用中提供了令人满意的学习能力。进行了广泛的实验,以评估和比较实际数据集上所提出方法的AUC(曲线下面积),准确性,召回率和训练时间。结果表明:(1)提出的直接从脑网络学习深度特征的方法优于浅层学习方法;(2)具有ELM增强结构的模型实现了更高的性能。本文探索具有深层功能和ELM的脑网络学习。结果表明,所提出的方法在AD检测的应用中提供了令人满意的学习能力。进行了广泛的实验,以评估和比较实际数据集上所提出方法的AUC(曲线下面积),准确性,召回率和训练时间。结果表明:(1)提出的直接从脑网络学习深度特征的方法优于浅层学习方法;(2)具有ELM增强结构的模型实现了更高的性能。本文探索具有深层功能和ELM的脑网络学习。结果表明,所提出的方法在AD检测的应用中提供了令人满意的学习能力。结果表明:(1)提出的直接从脑网络学习深度特征的方法优于浅层学习方法;(2)具有ELM增强结构的模型实现了更高的性能。本文探索具有深层功能和ELM的脑网络学习。结果表明,所提出的方法在AD检测的应用中提供了令人满意的学习能力。结果表明:(1)提出的直接从脑网络学习深度特征的方法优于浅层学习方法;(2)具有ELM增强结构的模型实现了更高的性能。本文探索具有深层功能和ELM的脑网络学习。结果表明,所提出的方法在AD检测的应用中提供了令人满意的学习能力。
更新日期:2019-11-29
down
wechat
bug