Alzheimer’s disease (AD) is one of the most common type of dementia, and there is no treatment to stop or reverse its progression so far. Since AD is one of the most leading cause of death these days and the cost of caring for this disease is expected to raise dramatically, early diagnosis is really crucial. This paper represents a Computer Aided Diagnosis system (CADs) for diagnosing Alzheimer’s disease. In this proposed method for decreasing heavy volumetric computations, 2D slices and consequently 2D convolutional neural network (CNN) is used and only half of the slices with higher quality are selected and half of them are removed. Also, for feature extraction and classification an improved AlexNet-SVM method is proposed which decreases computation volume and improve accuracy and efficiency. In this research, because of the considerable ability of PET images in showing body metabolism, these images are taken into account. The experimental results demonstrate the superiority of the offered model compared to existing models by improving performance and accuracy (Up to 96.39%) and decreasing the complex computations.

阿尔茨海默氏病（AD）是最常见的痴呆类型之一，到目前为止，尚无任何疗法可以阻止或逆转其发展。由于AD是近来最主要的死亡原因之一，并且护理该疾病的费用预计会急剧上升，因此早期诊断确实至关重要。本文介绍了一种用于诊断阿尔茨海默氏病的计算机辅助诊断系统（CAD）。在此提出的减少大量体积计算的方法中，使用了2D切片，因此使用了2D卷积神经网络（CNN），并且只选择了质量较高的切片中的一半，并去除了其中的一半。此外，对于特征提取和分类，提出了一种改进的AlexNet-SVM方法，该方法减少了计算量并提高了准确性和效率。在这项研究中 由于PET图像在显示人体新陈代谢方面具有相当大的能力，因此将这些图像考虑在内。实验结果通过提高性能和准确性（高达96.39％）并减少了复杂的计算，证明了所提供模型与现有模型相比的优越性。