Alzheimer disease is a neurodegenerative brain disorder leading to gradual loss of memory. Multiple biomarkers have been accepted for identifying the Alzheimer’s disease namely Neuroimaging, Cerebrospinal fluid proteins, blood and urine tests, genetic risk profilers. In this study, an extensive review has been done for Alzheimer disease prediction using diverse brain-imaging biomarkers through varied deep learning frameworks. A closer look revealed that taking into account multiple modalities of neuroimaging biomarkers always lead to better prediction accuracy for multi-class classification of Alzheimer disease. The paper further discusses about multiple open areas that need to be drilled down for establishing a model that can be accepted by medical community for Alzheimer prediction. This review work explores the different dimensions of neuroanatomical approach on which different deep learning frameworks that can be applied since the performance of designed model using 3-D subject-level, 3-D ROI-based and 3-D patch-level approaches varies. There is a need of extensive analysis for suitability of these methods for particular type of model.

阿尔茨海默氏病是导致记忆力逐渐丧失的神经退行性脑部疾病。多种生物标志物已被用于鉴定阿尔茨海默氏病，即神经影像学，脑脊液蛋白，血液和尿液检查，遗传风险分析仪。在这项研究中，通过各种深度学习框架，使用各种脑成像生物标记物对阿尔茨海默氏病的预测进行了广泛的综述。仔细观察后发现，考虑到神经影像生物标记物的多种方式，总能为阿尔茨海默氏病的多类别分类带来更好的预测准确性。本文进一步讨论了需要挖掘的多个开放区域，以建立可被医学界接受用于阿尔茨海默氏症预测的模型。这项审查工作探索了神经解剖学方法的不同维度，因为使用3-D主题级别，基于3-D ROI和3-D补丁程序级别的模型的设计模型的性能各不相同，因此可以应用不同的深度学习框架。需要针对这些方法对特定类型的模型的适用性进行广泛的分析。