摘要
为解决现有的轻度认知障碍计算机辅助诊断方法准确率较低、耗时较长的问题,对功能磁共振图像进行预处理,基于复杂网络理论构建脑网络,从中提取特征并采用LASSO方法特征选择,采用极限学习机实现轻度认知障碍的辅助诊断。此外,为了得到更合适的极限学习机分类模型,讨论了隐含层节点数与激活函数的选择对分类准确率的影响。结果显示:极限学习机分类器的分类准确率为93.3%,相比支持向量机和BP神经网络提高了13.3%和20.0%,而耗时相比支持向量机和BP神经网络分别减少了60.7%和99.5%,该方法较为明显地提高了轻度认知障碍辅助诊断的准确率和速度,对于轻度认知障碍辅助诊断的临床应用具有重要意义。
In order to solve the low-accuracy and time-consuming problems in current computer-aided diagnosis methods for mild cognitive impairment(MCI),this paper firstly preprocesses the rs-fMRI data and then builds brain networks based on complex network theory,from which the features are extracted and selected by the method of LASSO.Finally,this paper implements computer-aided diagnosis with extreme learning machine(ELM).In addition,in order to obtain a more suitable ELM classifier,this paper also discusses the influence of nodes in the hidden layer and the activation function on the classification accuracy.The result shows that the ELM classifier has a 93.3%accuracy,which is 13.3%and 20.0%higher than that of support vector machine(SVM)and BP neural network classifiers,respectively,while the time consumption is 60.7%and 99.5%shorter than the SVM and BP neural network.This method significantly improves the accuracy and speed of MCI assisted diagnosis,which is of great significance for promoting the clinical application of computer-aided MCI diagnosis.
作者
王之琼
蒋文静
刘秉佳
陈思冲
WANG Zhiqiong;JIANG Wenjing;LIU Bingjia;CHEN Sichong(College of Medicine and Biological Information Engineering, Northeastern University, Shenyang 110196, China)
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
北大核心
2021年第6期908-914,共7页
Journal of Harbin Engineering University
基金
国家自然科学基金项目(62072089)
中国博士后科学基金项目(2019T120216,2018M641705)
中央高校基本科研业务费(N2019007,N180101028,N180408019,N2024005-2)
国家级大学生创新创业训练计划(201910145101).
关键词
轻度认知障碍
阿尔兹海默病
极限学习机
支持向量机
BP神经网络
复杂网络理论
功能磁共振成像
计算机辅助诊断
mild cognitive impairment
Alzheimer′s disease
extreme learning machine
support vector machine
BP neural network
complex network theory
functional magnetic resonance imaging(fMRI)
computer-aided diagnosis
