基于迁移学习的多模态AD病程分类研究

Research on Multimodal Alzheimer Disease Course Classification Based on Transfer Learning

近年来,患阿尔茨海默病(Alzheimer’s Disease, AD)的人数逐年增加。临床研究显示,轻度认知障碍(Mild Cognitive Impairment, MCI)转化为AD的概率很大,因此,提高磁共振成像(Magnetic Resonance Imaging, MRI)和正电子发射断层扫描(Positron Emission Tomography, PET)等神经影像图对AD、 MCI的分类准确率十分必要。为了解决数据量少、标注困难的问题,首先使用CycleGAN网络对缺少的PET图进行生成;然后采用基于区域能量融合准则的小波变换算法对MRI图和PET图进行图像融合,能够极大程度的保留图像中的数据信息;最后利用迁移学习技术对轻量级网络MobileNet进行训练与微调。实验结果显示,在数据量较少的情况下,所提方法在泛化能力较好的同时,也获得了较高的准确率。

In recent years,the number of people suffering from Alzheimer’s Disease(AD)has increased year by year.Clinical research shows that the probability of mild cognitive impairment(MCI)transforming into AD is very high,so it is necessary to improve the classification accuracy of AD and MCI by magnetic resonance imaging(MRI),positron emission tomography(PET)and other neuroimaging images.In order to solve the problem of fewer data and difficulty in labeling,this paper first uses the CycleGAN network to generate the missing PET map;Secondly,the wavelet transform algorithm based on the regional energy fusion criterion is used to fuse MRI images and PET images,which can retain the data information in the images to a great extent;Finally,use the transfer learning technology to train and fine-tune the lightweight network MobileNet.The experimental results show that,in the case of fewer data,the method used in this paper has better generalization ability and higher accuracy.