Alzheimer’s disease (AD) is a leading cause of dementia, which causes serious health and socioeconomic problems. A progressive neurodegenerative disorder, Alzheimer’s causes the structural change in the brain, thereby affecting behavior, cognition, emotions, and memory. Numerous multivariate analysis algorithms have been used for classifying AD, distinguishing it from healthy controls (HC). Efficient early classification of AD and mild cognitive impairment (MCI) from HC is imperative as early preventive care could help to mitigate risk factors. Magnetic resonance imaging (MRI), a noninvasive biomarker, displays morphometric differences and cerebral structural changes. A novel approach for distinguishing AD from HC using dual-tree complex wavelet transforms (DTCWT), principal coefficients from the transaxial slices of MRI images, linear discriminant analysis, and twin support vector machine is proposed here. The prediction accuracy of the proposed method yielded up to 92.65 ± 1.18 over the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset, with a specificity of 92.19 ± 1.56 and sensitivity of 93.11 ± 1.29, and 96.68 ± 1.44 over the Open Access Series of Imaging Studies (OASIS) dataset, with a sensitivity of 97.72 ± 2.34 and specificity of 95.61 ± 1.67. The accuracy, sensitivity, and specificity achieved using the proposed method are comparable or superior to those obtained by various conventional AD prediction methods.
Alzheimer’s disease (AD) is the most familiar cause of dementia, with patients comprising 50%–80% of all dementia sufferers. The disease affects memory, cognition, and behavior. As AD is a neurodegenerative condition, several types of atrophy occur in the hippocampus and other areas of the brain. Despite being the 6th leading cause of death in the USA, it is not a common disease. Currently, there is no cure; however, some preventive measures can be taken to mitigate risk factors and slow the degenerative process. An estimated $605 billion globally and $220 billion in USA is spent annually on diagnosing AD. Many people suffer from AD worldwide, and demands on researchers are growing rapidly. MRI is an effective medical image construction technique, as it has the proven potential to view structural changes in the human brain, internal organs, and other tissues.
MRI produces high-quality structural images, providing distinctive tissue information, which enhances both the accuracy of brain pathology diagnosis and quality of treatment. A key advantage of this technique is its noninvasiveness. Many studies have been conducted using multivariate analysis algorithms and structural/functional MRI to classify neurological diseases [
The biomarkers used in our proposed method are MRI images from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and Open Access Series of Imaging Studies (OASIS) datasets. Our primary reason for using DTCWT over DWT is its effective representation of singularities (curves and lines), even though DWT has the advantage of representing the functions in multiscale and compressed forms. In DTCWT, shifts in magnitude variance can be achieved to a higher degree [
A total of 172 subjects from the ADNI dataset were used—86 AD and 86 HC. In addition, we used 95 subjects from the OASIS dataset—44 HC and 51 subjects suffering from very mild to mild AD.
Data used in the preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (
The ADNI was launched in 2003 as a public-private partnership led by Principal Investigator Michael W. Weiner, MD. The primary goal of the ADNI is to test whether serial MRI, positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of MCI and early-onset Alzheimer’s disease AD. For up-to-date information, visit
Summary of subject’s demographics status.
AD | Normal | |
---|---|---|
Number of subjects | 86 | 86 |
43 males | 46 males | |
43 females | 40 females | |
Average age | 77.30 | 76.05 |
Average education points | 14.65 | 15.93 |
MMSE | 23.48 | 29.08 |
In addition, we utilized MRI images downloaded from the OASIS dataset. OASIS is a database designed to compile MRI datasets and make them freely accessible to the scientific community. OASIS compiles two types of data: cross-sectional MRI data and longitudinal MRI data. Our study utilized cross-sectional MRI data, as our aims are to develop an automatic system for detecting AD, for which longitudinal MRI data is not optimal.
The OASIS dataset consists of 416 subjects aged between 18 and 96 years. Our study included 51 AD patients (35 with CDR = 0.5 and 16 with CDR = 1) out of 100 having dementia and 44 HC out of 98 normal subjects. Table
Statistical OASIS data details used in our learning.
Factors | Normal | Very mild & mild AD |
---|---|---|
Number of patients | 44 | 51 |
Age | 84.40 (76–96) | 82.11 (76–96) |
Education | 3.34 (1–5) | 3.13 (1–5) |
Socioeconomic status | 2.31 (1–5) | 2.82 (1–5) |
CDR (0.5/1) | 0 | 35/16 |
MMSE | 28.72 (25–30) | 24.82 (18–30) |
Clinical dementia scale.
CDR | Rank |
---|---|
0.5 | Very mild dementia |
1 | Mild |
2 | Moderate |
3 | Severe |
The proposed approach is made up of 4 phases: preprocessing and slice extraction, feature extraction, projection of features into lower dimension, and efficient classification of the disease. Figure
Flowchart of DTCWT-based classification performance of AD from HC.
All MRI images used for training and testing the TSVM of our proposed approach are viewed using the ONIS toolbox and exported as 2D MRI image slices. All images are in PNG format, and the dimensions of OASIS image slices are 176 × 208; the dimensions of the ADNI image slices are 256 × 166. The range of selection of those slices was performed manually from the tissue center for information clarity. The images are resized to 256 × 256 for further processing. A sample of a brain image slice is depicted in Figure
MR image slice sample (axial slice view after preprocessing).
Normal
Alzheimer
Wavelet transform (WT) is one of the most frequently used feature extraction techniques for MR images. For our proposed approach, we extract the DTCWT [
Block diagram for a 3-level DTCWT.
The DTCWT can be denoted in matrix form as
For the input image
The DTCWT coefficients of input images are shift invariant; they do not change when an image is shifted in time or space. In addition, DTCWT employs segregation of 6 diverse directions (±15, ±30, and ±45) for 2D images and 28 different directions for 3D images, while conventional DWT only allows for isolation of horizontal and vertical directions. For each 2D slice subject image, we extracted 5-level DTCWT coefficients from one scale.
Principal component analysis (PCA) [
PCA implementation for feature reduction.
The PCA is summarized as follows:
Calculating the mean of the data and zero mean data Constructing the covariance matrix Acquiring the eigenvalue and the eigenvector Projecting the data matrix with eigenvectors corresponding to the highest to lowest eigenvalues.
A generalized Fisher linear discriminant [
To find the class separation projection axis, it is necessary to determine between-class scatter and within-class variability.
The between class variable matrix can be denominated by sample variance as
Within class variance matrix can be expressed as
The generalized Rayleigh coefficient is
If
The PCA coefficients can be projected onto
The final feature matrix
Jayadeva and Chandra [
Mathematically, the TSVM primal problem can be optimized by solving the following two quadratic programming problems:
Here,
In this article, our proposed approach is presented using Fisher linear discriminant analysis of DTCWT principal components. The details of our proposed method are shown in Figure
However, there are multiple drawbacks to conventional wavelet transform. These include drift in wavelet coefficient oscillation towards positive and negative around singularities, shift variance of signal (which may cause oscillation of wavelet coefficient samples around singularities), substantial aliasing of amply spaced wavelet coefficient patterns, and lack of directional selectivity perturbs to process and model geometric image features (such as edges and ridges). In these cases, flaws regarding conventional DWT are not experienced by Fourier transform. Inspired by Fourier transform, our improved DTCWT is used to overcome these drawbacks. Previous studies have shown that DTCWT feature-based AD disease detection performs better than typical DWT-based feature extraction [
Misclassification rates and higher dimensionality of features present problems concerning pattern classification. For smooth classification, dimensionality reduction techniques are employed to transform data from higher to lower dimensional spaces. PCA is the most frequently applied linear transformation and addresses these concerns. Extracted features are analyzed using PCA for feature reduction. For each MRI image from the OASIS and ADNI datasets, there are 49,152 (1536 × 32) features. After applying PCA, this is reduced to 95 × 94 for OASIS data and 172 × 171 for ADNI data.
After PCA, the classification may still not be sufficient, as PCA does not account for variability of features within a class or between classes. To ensure that the PCs are more separable, it is needed to transform data onto another space combining directions that will find axes, which will maximize the gap between different classes. Thus, LDA is applied to project PCs onto new projection axes for more effective disease classification.
TSVM is an emerging efficient pattern classification and regression algorithm in machine learning. Numerous studies have shown that TSVM is highly effective in terms of classification, regression performance, and time complexity [
All programs are executed in MATLAB 2015b installed on an Intel (R) Core (TM) i3-4160 CPU system. The time complexity of the extraction of DTCWT and DWT coefficients from a 2D MRI image slice are 0.5148 and 0.5109, respectively. There is no significant difference in CPU-elapsed time when comparing transform methods. As a dimensionality reduction technique, we used PCA to omit higher dimensional input features.
In addition, it is not feasible to train and test a classifier with higher dimensional features due to elapsed time. The CPU-elapsed time to achieve TSVM classification performance was approximately 88.40 seconds without reducing dimensions. The time required for our proposed method is approximately 15.74 seconds—faster than the methods that do not employ fisher discriminant analysis.
The performance of a binary classifier can be visualized using a confusion matrix, as shown in Table
Confusion matrix for a binary classifier to distinguish between two classes (S1 and S2).
True class | Predicted class | |
---|---|---|
S1 (patients) | S2 (controls) | |
S1 (patients) | TP | FN |
S2 (controls) | FP | TN |
Accuracy is determined measuring the proportion of examples that are correctly labeled by a classifier:
This may not be an ideal performance metric if the class distribution of the dataset is unbalanced.
For example, if class
Sensitivity measures the proportion of correctly identified patients, and specificity measures the proportion of correctly identified controls. Additionally, some other frequently used statistical performance evaluation measures such as
These measures are defined as
The previous measures are likely to provide an efficient overall performance assessment of a classifier.
In this study, the proposed hybrid method has been used for OASIS and ADNI data to distinguish control subjects from AD subjects. The recorded classification performance regarding accuracy (acc), sensitivity (sens), and specificity (spec) has been shown in a bar diagram in Figure
Bar chart of DTCWT-based classification performance of AD from HC over ADNI dataset.
Bar chart of DTCWT-based classification performance of AD from HC over OASIS dataset.
The number of principal components versus classification performance graph of proposed method.
The accuracies, sensitivities, specificities, and other statistical performance measures obtained with 10–20 runs of 10-fold SCV and 5-fold SCV are shown in Tables
Performance evaluation over ADNI dataset.
Methods | Accuracy | Sensitivity | Specificity | Precision | Recall | f_measure | gmean |
---|---|---|---|---|---|---|---|
Proposed | 92.65 ± 1.18 | 93.11 ± 1.29 | 92.19 ± 1.56 | 92.78 ± 1.27 | 93.11 ± 1.29 | 92.63 ± 1.19 | 92.46 ± 1.24 |
DTCWT+PCA+TSVM | 91.77 ± 0.85 | 92.48 ± 0.89 | 91.13 ± 1.31 | 91.73 ± 0.95 | 92.48 ± 0.89 | 91.72 ± 0.77 | 91.57 ± 0.91 |
Performance evaluation over OASIS dataset.
Methods | Accuracy | Sensitivity | Specificity | Precision | Recall | f_measure | gmean |
---|---|---|---|---|---|---|---|
Proposed | 96.68 ± 1.44 | 97.72 ± 2.34 | 95.61 ± 1.67 | 96.13 ± 1.57 | 97.72 ± 2.34 | 96.76 ± 1.51 | 96.56 ± 1.44 |
DTCWT+PCA+TSVM | 95.46 ± 1.35 | 97.55 ± 1.26 | 93.36 ± 2.39 | 94.14 ± 2.01 | 97.55 ± 1.26 | 95.61 ± 1.28 | 95.29 ± 1.42 |
Although comparison with conventional methods can be difficult, we have compared our approach with some recent conventional disease detection algorithms using both datasets.
To analyze the performance over the ADNI dataset, the classification performance has been documented with both run-wise fold-wise classification, as shown in Tables
Run- and fold-wise classification performance of proposed approach over ADNI dataset.
Folds | Runs | Run 1 | Run 2 | Run 3 | Run 4 | Run 5 | Run 6 | Run 7 | Run 8 | Run 9 | Run 10 | |
---|---|---|---|---|---|---|---|---|---|---|---|---|
Fold 1 | 94.44 | 100 | 100 | 100 | 94.44 | 88.8889 | 100 | 87.5 | 94.44 | 100 |
| |
Fold 2 | 100 | 94.117 | 100 | 88.23 | 82.35 | 94.11 | 81.25 | 94.11 | 94.44 | 88.88 | ||
Fold 3 | 94.117 | 94.117 | 94.11 | 88.88 | 100 | 82.35 | 100 | 100 | 88.88 | 88.23 | ||
Fold 4 | 94.117 | 88.235 | 88.88 | 94.11 | 100 | 93.75 | 94.117 | 94.11 | 100 | 82.35 | ||
Fold 5 | 87.5 | 88.888 | 88.88 | 94.11 | 88.23 | 100 | 93.75 | 100 | 87.5 | 94.44 | ||
Fold 6 | 100 | 94.117 | 87.5 | 88.88 | 100 | 76.47 | 88.23 | 77.77 | 94.11 | 94.44 | ||
Fold 7 | 87.5 | 94.117 | 87.5 | 93.75 | 100 | 83.33 | 100 | 94.11 | 82.35 | 93.75 | ||
Fold 8 | 87.5 | 100 | 100 | 88.88 | 100 | 94.44 | 100 | 83.33 | 87.5 | 94.11 | ||
Fold 9 | 94.444 | 100 | 94.44 | 94.11 | 88.88 | 94.11 | 100 | 88.235 | 88.888 | 87.5 | ||
Fold 10 | 94.444 | 83.333 | 83.33 | 94.11 | 82.35 | 100 | 88.88 | 94.117 | 100 | 100 | ||
Fold-wise accuracy | 93.406 | 93.692 | 92.46 | 92.512 | 93.62 | 90.747 | 94.624 | 91.3317 | 91.813 | 92.37 |
Run- and fold-wise classification performance of the DTCWT + PCA + TSVM method over ADNI dataset.
Folds | Runs | Run 1 | Run 2 | Run 3 | Run 4 | Run 5 | Run 6 | Run 7 | Run 8 | Run 9 | Run 10 | |
---|---|---|---|---|---|---|---|---|---|---|---|---|
Fold 1 | 88.88 | 94.11 | 94.11 | 87.5 | 87.5 | 77.77 | 75 | 100 | 82.35 | 88.88 |
| |
Fold 2 | 94.11 | 100 | 100 | 94.44 | 100 | 94.44 | 88.88 | 94.11 | 88.23 | 87.5 | ||
Fold 3 | 94.11 | 87.5 | 88.23 | 93.75 | 94.11 | 94.11 | 88.88 | 88.23 | 93.75 | 76.47 | ||
Fold 4 | 93.75 | 82.35 | 88.23 | 88.23 | 100 | 94.11 | 94.11 | 88.23 | 76.47 | 100 | ||
Fold 5 | 88.88 | 94.11 | 94.11 | 83.33 | 82.35 | 94.11 | 82.35 | 88.88 | 94.44 | 100 | ||
Fold 6 | 94.11 | 82.35 | 94.44 | 100 | 100 | 100 | 87.5 | 94.11 | 88.88 | 88.88 | ||
Fold 7 | 83.33 | 94.44 | 100 | 100 | 83.33 | 87.5 | 100 | 88.23 | 100 | 100 | ||
Fold 8 | 87.5 | 94.44 | 83.33 | 82.35 | 88.23 | 93.75 | 94.44 | 88.23 | 93.75 | 83.33 | ||
Fold 9 | 94.44 | 100 | 94.44 | 88.88 | 100 | 88.23 | 100 | 82.35 | 88.88 | 100 | ||
Fold 10 | 94.11 | 94.11 | 88.23 | 88.23 | 94.11 | 100 | 100 | 100 | 100 | 100 | ||
Fold-wise accuracy | 91.32 | 92.34 | 92.51 | 90.67 | 92.96 | 92.40 | 91.11 | 91.24 | 90.67 | 92.50 |
We have compared several recently used sets of algorithms and methods [
Classification performance of AD from HC over ADNI data.
Methods | Accuracy | Sensitivity | Specificity |
---|---|---|---|
Proposed | 92.65 ± 1.18 | 93.11 ± 1.29 | 92.19 ± 1.56 |
DTCWT + PCA + TSVM | 91.77 ± 0.85 | 92.48 ± 0.89 | 91.13 ± 1.31 |
DTCWT + PCA + LDA + Kernel SVM | 90.181 ± 0.97 | 90.276 ± 1.60 | 90.101 ± 1.23 |
DTCWT + PCA + Kernel SVM | 82.74 ± 1.24 | 84.43 ± 1.51 | 81.18 ± 1.85 |
DWT + PCA + LDA + TSVM | 86.75 ± 1.69 | 89.32 ± 1.43 | 84.23 ± 2.21 |
DWT + PCA + TSVM | 85.88 ± 1.16 | 88.93 ± 1.61 | 88.93 ± 2.02 |
DTCWT + PCA + LDA + ANN | 86.97 ± 1.30 | 86.25 ± 1.78 | 87.72 ± 3.51 |
DTCWT + PCA + LDA + KNN | 83.89 ± 0.75 | 81.41 ± 1.33 | 86.34 ± 1.08 |
DTCWT + PCA + LDA + AdaBoost (tree) | 84.48 | 83.72 | 85.26 |
DWT + PCA + ANN [ |
80.05 ± 0.72 | 81.538 ± 1.41 | 78.974 ± 1.09 |
DWT + PCA + KNN [ |
79.964 ± 1.19 | 78.771 ± 2.37 | 81.08 ± 1.67 |
[ |
85 | 82 | 88 |
Likewise, to analyze and stratify OASIS dataset, identical methods have been used, namely run-wise and fold-wise classifications, as depicted in Tables
Run- and fold-wise classification performance of the proposed approach over OASIS dataset.
Folds | Runs | Run 1 | Run 2 | Run 3 | Run 4 | Run 5 | Run 6 | Run 7 | Run 8 | Run 9 | Run 10 | |
---|---|---|---|---|---|---|---|---|---|---|---|---|
Fold 1 | 94.44 | 94.44 | 100 | 88.23 | 100 | 100 | 100 | 94.11 | 100 | 100 |
| |
Fold 2 | 94.11 | 100 | 88.23 | 88.88 | 88.88 | 100 | 94.11 | 100 | 88.88 | 93.75 | ||
Fold 3 | 94.44 | 94.11 | 100 | 94.11 | 100 | 94.44 | 100 | 94.11 | 100 | 100 | ||
Fold 4 | 100 | 100 | 100 | 94.44 | 100 | 94.44 | 100 | 100 | 100 | 100 | ||
Fold 5 | 100 | 88.88 | 100 | 100 | 94.44 | 100 | 94.11 | 100 | 94.11 | 94.44 | ||
Fold-wise accuracy | 96.60 | 95.49 | 97.64 | 93.13 | 96.66 | 97.77 | 97.64 | 97.64 | 96.60 | 97.63 |
Run- and fold-wise classification performance of the DTCWT + PCA + TSVM method over OASIS dataset.
Folds | Runs | Run 1 | Run 2 | Run 3 | Run 4 | Run 5 | Run 6 | Run 7 | Run 8 | Run 9 | Run 10 | |
---|---|---|---|---|---|---|---|---|---|---|---|---|
Fold 1 | 100 | 94.11 | 94.44 | 100 | 94.44 | 94.44 | 94.11 | 94.44 | 88.8 | 94.44 |
| |
Fold 2 | 100 | 100 | 88.23 | 94.44 | 94.44 | 100 | 94.11 | 94.44 | 100 | 94.44 | ||
Fold 3 | 94.44 | 83.33 | 94.44 | 94.11 | 100 | 94.44 | 94.44 | 94.11 | 94.44 | 100 | ||
Fold 4 | 94.44 | 100 | 94.44 | 94.11 | 94.44 | 88.88 | 100 | 100 | 100 | 88.88 | ||
Fold 5 | 94.11 | 100 | 94.11 | 94.44 | 100 | 87.5 | 100 | 94.44 | 100 | 100 | ||
Fold-wise accuracy | 96.60 | 95.49 | 93.13 | 95.42 | 96.66 | 93.05 | 96.53 | 95.49 | 96.66 | 95.55 |
We observed, as shown in Tables
Algorithm performance comparison over OASIS MRI data.
Algorithm | Accuracy | Sensitivity | Specificity | Precision |
---|---|---|---|---|
Proposed |
|
|
|
|
DTCWT + PCA + TSVM | 95.46 ± 1.35 | 97.55 ± 1.26 | 93.36 ± 2.39 | 94.15 ± 2.01 |
DWT + PCA + LDA + TSVM | 87.23 ± 1.65 | 89.61 ± 2.25 | 84.85 ± 1.66 | 86.66 ± 1.99 |
DWT + PCA + TSVM | 86.19 ± 1.50 | 88.83 ± 1.98 | 83.5 ± 1.87 | 85.66 ± 1.84 |
DTCWT + PCA + LDA + ANN | 88.59 + 2.08 | 88.75 + 2.75 | 89.55 + 3.96 | NA |
DTCWT + PCA + LDA + KNN | 83.69 + 1.57 | 85.7 + 1.94 | 81.8 + 1.45 | NA |
DTCWT + PCA + LDA + AdaBoost (tree) | 87.45 | 88.59 | 86.26 | NA |
BRC + IG + SVM [ |
90.00 (77.41, 96.26) | 96.88 (82.01, 99.84) | 77.78 (51.92, 92.63) | NA |
BRC + IG + Bayes [ |
92.00 (79.89, 97.41) | 93.75 (77.78, 98.27) | 88.89 (63.93, 98.05) | NA |
BRC + IG + VFI [ |
78.00 (63.67, 88.01) | 65.63 (46.78, 80.83) | 100.00 (78.12, 100) | NA |
MGM + PEC + SVM [ |
92.07 ± 1.12 | 86.67 ± 4.71 | N/A | 95.83 ± 5.89 |
GEODAN + BD + SVM [ |
92.09 ± 2.60 | 80.00 ± 4.00 | NA | 88.09 ± 5.33 |
TJM + WTT + SVM [ |
92.83 ± 0.91 | 86.33 ± 3.73 | N/A | 85.62 ± 0.85 |
VBM + RF [ |
89.0 ± 0.7 | 87.9 ± 1.2 | 90.0 ± 1.1 | NA |
DF + PCA + SVM [ |
88.27 ± 1.9 | 84.93 ± 1.21 | 89.21 ± 1.6 | 69.30 ± 1.91 |
EB + WTT + SVM + RBF [ |
86.71 ± 1.93 | 85.71 ± 1.91 | 86.99 ± 2.30 | 66.12 ± 4.16 |
EB + WTT + SVM + Pol [ |
92.36 ± 0.94 | 83.48 ± 3.27 | 94.90 ± 1.09 | 82.28 ± 2.78 |
Curvelet + PCA + KNN [ |
89.47 | 94.12 | 84.09 | NA |
US + SVDPCA + SVM-DT [ |
90 | 94 | 71 | NA |
To further verify the efficacy of the proposed method, we compared it with 12 state-of-the-art approaches, as shown in Table
The results show that US + SVD-PCA + SVM-DT [
Similarly, BRC + IG + VFI [
All other methods achieved satisfying results. VBM + RF [
DF + PCA + SVM [
EB + WTT + SVM + RBF [
In addition, MGM + PEC + SVM [
Finally, taking classification performance into consideration, our approach outperforms all other methods analyzed here. We have also produced promising performance metrics for sensitivity and specificity. Hence, we submit that our results are either superior or comparable to the other compared methods.
Our proposed experiment uses LDA on the principal components of DTCWT coefficients and TSVM to stratify AD. Our proposed detection method for the ADNI dataset yielded an accuracy of 92.65 ± 1.18% with high sensitivity and specificity. Our proposed method also outperforms those of Zhang et al. [
In the future, we will carry forward our research focusing on the following: (i) 3D DTCWT-based feature extraction with multiresolution analysis and classification and (ii) convolutional neural network- (CNN-) based classification using 3D MRI.
The investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in the analysis or writing of this report.
The authors declare that they have no conflicts of interest.
This research was supported by the Brain Research Program through the National Research Foundation of Korea funded by the Ministry of Science, ICT & Future Planning (NRF-2014M3C7A1046050). And this study was supported by the research funds from Chosun University, 2017. Data collection and sharing for this project was funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant no. U01 AG024904) and DOD ADNI (Department of Defense Award no. W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging and the National Institute of Biomedical Imaging and Bioengineering and through generous contributions from the following: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech; BioClinica Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd. and its affiliated company Genentech Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development LLC; Johnson & Johnson Pharmaceutical Research & Development LLC; Lumosity; Lundbeck; Merck & Co. Inc.; Meso Scale Diagnostics LLC; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research provide funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (