基于深度神经网络的阿尔茨海默病早期诊断算法

An Algorithm for Early Diagnosis of Alzheimer's Disease Based on Deep Neural Network

关键词:  阿尔茨海默病  神经退行性疾病  疾病诊断  深度学习  对比学习  多层感知机

Key words:  Alzheimer's disease  neurodegenerative disease  disease diagnosis  deep learning  contrastive learning  multilayer perceptron

摘要:

由于阿尔茨海默病(Alzheimer's Disease，AD)目前无法治愈，只能通过早期干预的方式缓解其恶化，因此使用神经影像学技术对阿尔茨海默病进行早期诊断已经在临床中得到广泛应用。本文利用深度学习技术，对基于磁共振成像(Magnetic Resonance Imaging，MRI)数据的阿尔茨海默病早期诊断进行了研究。首先，引入样本重加权策略来缓解疾病诊断数据集中普遍存在的类不平衡问题对模型的影响。其次，设计监督对比学习来提升模型的特征提取能力。在4个神经退行性疾病诊断数据集上的消融实验结果表明，本文提出的方法取得了比现有算法更好的性能。

Abstract:

Alzheimer's Disease (AD) is currently incurable and its deterioration can only be mitigated by early intervention.Therefore,early diagnosis of Alzheimer's disease by neuroimaging has been widely used in clinical practice.A deep learning network model is built based on the Magnetic Resonance Imaging (MRI) data for the early diagnosis of Alzheimer's disease.First,a sample reweighting strategy is introduced to mitigate the impact of the class imbalance problem prevalent in disease diagnosis datasets on the model.Second,supervised contrastive learning is designed to improve the feature extraction capability of the model.The experimental results on four neurodegenerative disease diagnosis datasets show that the method proposed in this paper achieves better performance than existing methods.