波谱学杂志 ›› 2020, Vol. 37 ›› Issue (3): 321-331.doi: 10.11938/cjmr20192769

• 研究论文 • 上一篇    下一篇

基于磁共振图像和改进的UNet++模型区分阿尔茨海默症患者和健康人群

赵尚义, 王远军   

  1. 上海理工大学 医学影像工程研究所, 上海 200093
  • 收稿日期:2019-07-16 出版日期:2020-09-05 发布日期:2019-08-27
  • 通讯作者: 王远军,Tel:13761603606,E-mail:yjusst@126.com. E-mail:yjusst@126.com
  • 基金资助:
    上海市自然科学基金资助项目(18ZR1426900).

Classification of Alzheimer's Disease Patients Based on Magnetic Resonance Images and an Improved UNet++ Model

ZHAO Shang-yi, WANG Yuan-jun   

  1. Institute of Medical Imaging Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China
  • Received:2019-07-16 Online:2020-09-05 Published:2019-08-27

摘要: 阿尔茨海默症(Alzheimer's Disease,AD)是一种神经退行性疾病,高效准确的早期诊断对其治疗至关重要.本文提出了一种融合多语义级别的深度卷积神经网络结构,基于磁共振图像,用于区分AD患者与正常受试者的方法.首先,在传统UNet++网络的基础上改进了深度监督整合算法;然后,构建了一种新的特征融合结构,进一步细化了不同语义级别的特征;最后,基于不同组织区域(白质、灰质和脑脊液)的磁共振图像,使用本文所提出的方法区分AD患者和正常受试者,并探究了从不同组织获得的信息对分类准确率的影响.实验结果表明,使用本文提出的方法区分两类人群的最高准确率为98.74%,平均准确率为98.47%,高于目前文献报道的其他方法.

关键词: 磁共振成像(MRI), 深度学习, 特征融合, 阿尔茨海默症

Abstract: Alzheimer's disease (AD) is one of the most common forms of dementia and a degenerative mental disorder that seriously affects people's daily lives. Rapid and effective diagnosis is essential for the treatment of patients with Alzheimer's disease. To solve this problem, this paper proposes a deep convolutional neural network structure with multiple semantic levels to classify AD patients and healthy controls from magnetic resonance imaging (MRI) data. Firstly, the deep supervision integration algorithm and the classification model of Alzheimer's disease based on the traditional UNet++ network were improved. Then, a new feature fusion structure was constructed, which further refined the different semantic levels. Lastly, the proposed protocol was applied to different tissue regions (e.g., white matter, gray matter and cerebrospinal fluid), and the effects of different tissue information combinations on the classification outcome were explored. The method proposed was applied to the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset to classify the AD patients. The results demonstrated that the highest accuracy of 98.74%, and an average accuracy of 98.47%.

Key words: magnetic resonance imaging (MRI), deep learning, feature fusion, Alzheimer's disease

中图分类号: