Abstract

Convolutional neural networks (CNNs) have been widely applied for brain disease detection based on neuroimaging data. It shows significant and accurate results. Alzheimer's disease is one of the most common neurodegenerative disorders affecting a person's ability and cannot be reversed. The early detection of this disease can slow its progression. Researchers adopt Magnetic Resonance Imaging (MRI) modality as a biomarker for Alzheimer's disease detection based on CNN. Building a CNN model that enhances reliable feature extraction and representation is challenging for Alzheimer's disease detection. The main drawback of CNN is that increasing model size could affect the model's performance. Furthermore, CNN cannot capture long relationships between image features. In this paper, we evaluated different machine learning algorithms for AD diagnosis. In addition, we proposed a Depthwise Separable Convolutional ResNet with an attention mechanism to address these challenges. The proposed network replaces the standard convolution operation with a depth-wise separable convolution to decrease the network parameters and size, which reduces overfitting. The attention mechanism is employed to boot the network feature representation capability. We evaluated our model using MRI data from the Open Access Series of Imaging Studies (OASIS) dataset. The proposed model outperforms different CNN models reaching an accuracy of 99%. We evaluated the proposed method for Multimodal binary classification of Alzheimer's disease and we obtained an accuracy of 92%.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call