Using Three-dimensional Convolutional Neural Networks for Alzheimer’s Disease Diagnosis

Abstract

Alzheimer's disease (AD) is an irreversible neurodegenerative disease. Pathology shows atrophy of brain tissue, senile plaques, neurofibrillary tangles, and so forth. Magnetic resonance imaging (MRI) is the most sensitive brain imaging method in a clinic, which provides detailed anatomical structure information of the brain and is commonly studied with pattern recognition methods for AD diagnosis. Most existing methods extract hand-crafted imaging features or brain region-of-interest images to train a classifier to recognize AD. In this study, a threedimensional convolutional neural network (3D-CNN) is implemented to detect AD. The proposed network is trained with 3D magnetic resonance (MR) images to extract the spatial features. A uniform experimental design is used to optimize the network parameters and improve the 3D-CNN performance. To ensure satisfactory performance on a small amount of training data, a transfer learning technology is proposed to improve the recognition rate of the 3D-CNN. In addition, the uniform experimental design (UED) method is used to determine the optimal parameter combination of the network and improve the 3D-CNN performance. The validation data in this study are from the Open Access Series of Imaging Studies (OASIS), where the OASIS-1 data set is used as the original data set and the 3D-CNN is trained as the pretraining model. Experimental results show that when 10, 30, 60, and 90% of the OASIS-2 data set are used to train the pre-trained 3D-CNN, the average accuracy reaches 74.66, 86.99, 94.58, and 97.02%, respectively. In addition, compared with the original manual design parameters, the proposed 3D-CNN with the best parameter combination improves the recognition rate by 2.07 percentage points.

An MRI Scans-Based Alzheimer’s Disease Detection via Convolutional Neural Network and Transfer Learning

“…The remaining works [15][16][17][18] formulated the problem as binary AD detection, where the detector only determines if the participant suffers from AD (without the information of the severity); • Features and algorithms: Works [15,17] separated the feature extraction and AD detection into two parts using two algorithms. Our work and [16,18] formulated the feature extraction and AD detection with one algorithm; • Type of cross-validation: Work [13] did not employ cross-validation, which may result in insufficiency in hyperparameter tuning and evaluation of the model overfitting. Threefold cross-validation was adopted in our work, whereas fivefold cross-validation was used in [12,14].…”

“…The discussion of the methodology and results of the existing works is separated based on each dataset. The works are [ 11 , 12 , 13 , 14 ] for OASIS-1, [ 15 , 16 , 17 , 18 ] for OASIS-2, and [ 19 , 20 , 21 , 22 ] for OASIS-3.…”

“…Results revealed that the model achieved a sensitivity of 88.2% and a specificity of 100%. A subset of OASIS-2 was selected in [ 16 ] to build a binary AD detection model using 3D convolutional neural network (CNN). The accuracy of the model was reported as 97%.…”

“…Only a portion of the dataset was considered in the model implementation and performance analysis in works [ 13 , 14 , 16 , 18 , 19 ];…”

“…Only a portion of the dataset was considered in the model implementation and performance analysis in works [13,14,16,18,19]; • Some of the existing works did not employ cross-validation [13,[15][16][17][18][19][20][21][22] and improperly defined the ratio of cross-validation [11] in the performance evaluation and analysis of the AD detection models. The trained models without cross-validation may not be designed with optimal sets of hyperparameters and may be more prone to model overfitting;…”