Research Article  Open Access
Debesh Jha, JiIn Kim, GooRak Kwon, "Diagnosis of Alzheimer’s Disease Using DualTree Complex Wavelet Transform, PCA, and FeedForward Neural Network", Journal of Healthcare Engineering, vol. 2017, Article ID 9060124, 13 pages, 2017. https://doi.org/10.1155/2017/9060124
Diagnosis of Alzheimer’s Disease Using DualTree Complex Wavelet Transform, PCA, and FeedForward Neural Network
Abstract
Background. Errorfree diagnosis of Alzheimer’s disease (AD) from healthy control (HC) patients at an early stage of the disease is a major concern, because information about the condition’s severity and developmental risks present allows AD sufferer to take precautionary measures before irreversible brain damage occurs. Recently, there has been great interest in computeraided diagnosis in magnetic resonance image (MRI) classification. However, distinguishing between Alzheimer’s brain data and healthy brain data in older adults (age > 60) is challenging because of their highly similar brain patterns and image intensities. Recently, cuttingedge feature extraction technologies have found extensive application in numerous fields, including medical image analysis. Here, we propose a dualtree complex wavelet transform (DTCWT) for extracting features from an image. The dimensionality of feature vector is reduced by using principal component analysis (PCA). The reduced feature vector is sent to feedforward neural network (FNN) to distinguish AD and HC from the input MR images. These proposed and implemented pipelines, which demonstrate improvements in classification output when compared to that of recent studies, resulted in high and reproducible accuracy rates of 90.06 ± 0.01% with a sensitivity of 92.00 ± 0.04%, a specificity of 87.78 ± 0.04%, and a precision of 89.6 ± 0.03% with 10fold crossvalidation.
1. Introduction
Alzheimer’s disease (AD) is an irremediable neurodegenerative disorder that causes dementia in elderly people around the globe. It has been predicted that the pervasiveness of AD will double within the next 2 decades and that one out of every 85 people will be afflicted with the disease by 2050 [1]. Therefore, there is a need to identify neuroimaging biomarkers that can grant accurate and early diagnoses of dementia. In addition, to diagnose an AD sufferer clinically at a primitive disease stage, many imaging biomarkers must be identified using different imaging modalities, such as MRI [2], position emission tomography (PET) [3], functional magnetic resonance imaging (fMRI) [4], singlephoton emission computed tomography (SPECT) [5], and magnetic resonance spectral imaging (MRSI) [6].
An accurate and early diagnosis of AD and identification of the risk of progression from mild cognitive impairment (MCI) to AD provide AD sufferers with awareness of the condition’s severity and allow them to take preventative measures, such as making lifestyle changes and taking medications [7]. Currently, many neurologists and medical analysts have been spending significant time to researching technique to allow for early diagnosis of AD, and encouraging results have been frequently achieved [8]. MRI is an influential, noninvasive brain imaging technique that provides higherquality information about the shape and volume of the brain than computed tomography (CT), SPECT, and PET scans. It provides superior soft tissue differentiation, high spatial resolution, and better contrast and can even identify tiny irregularities in the brain [9]. Moreover, the diagnostic use of MRI has been tremendously improved due to the automated and precise labeling of MR images, which performs an important role in identifying AD in related patients from healthy and elderly controls (HC) [10].
Earlier, majority of diagnosis work was accomplished manually or semimanually for measuring a priori region of interest (ROI) of MRI, based on the reality that subjects with AD experience have more cerebral atrophy when compared to HCs [11, 12]. Most of this ROIbased examination focused on the contracting of the cortex and hippocampus and amplified ventricles. Nevertheless, ROIbased approaches are not practicable in hospitals because of few shortcomings: (i) ROI technique needs a priori data and expert knowledge. (ii) The manual diagnosis accuracy is dependent on the knowledge of physicians and interpreter [13]. (iii) The interaction among the voxels was troublesome to enforce. (iv) It was essential to explore other potential areas that may be linked to AD [14]. (v) Automatic segmentation of ROI was not beneficial in practice, and investigator needed to segment the brain using hand [15]. Therefore, automated methods can assist physician in diagnosing diseases from images such as those produced by MRI, for which many slices are extracted from the tissues and long periods of may be necessary for the evaluation of the images.
The aim of this article is to present an automated approach for diagnosing AD by using the “whole brain analysis” method. It has achieved popularity, since it examines entire voxels of the brain. It is not essential to segment the brain as earlier, and it does not require any biomarker for the classification purpose. The main drawback is dimensionality that can be resolved through highspeed computers, which is comparably inexpensive [16]. The wholebrain investigation laboriously relies on true computation, and it can only be finished by a computer researcher after a physician assisted in labeling the input data as either AD or HC. Usually, the wholebrain inspection labels the entire brain as a ROI, where two stages are involved, namely, feature extraction and classification.
Scholars have presented different methods to extract effective features for the detection of AD and other types of pathological brain disease. Additionally, classification models and methods survive; nevertheless, not all of them are suitable for the processing of MR brain images. Based on latest literature, we found two drawbacks with the previous work: (i) The discrete wavelet transform (DWT) is usually utilized for feature extraction. The DWT has better directional selectivity in horizontal, vertical, and diagonal directions and has better image representation than Fourier transform, but its major drawbacks are that it has poor directionality, is sensitive to shifts, and lacks phase information. (ii) Most of the stateoftheart mechanisms consider only single slicebased detection (SSD) per patient. The obtained slices may not contain the foci of the disease.
To tackle above problems, we suggested two improvements. First, we propose a DTCWT that possesses attractive properties for image processing, including shift invariance and high directionality. Second, we consider multiple slices for each patient unlike previous studies, so that information gain is more consistent, reliable, and accurate. In hospitals, multiple slicebased detection is utilized because of its inexpensiveness. Research has clearly showed that the DTCWT is more suitable than the traditional wavelet domain for feature extraction [17].
Our contribution aims to introduce a novel method for AD detection with higher accuracy than stateoftheart methods, on the basis of DTCWT, PCA, and ANN technique. Furthermore, we build a computeraided diagnosis (CAD) system, which can be utilized in the early diagnosis of ADrelated brain area and subjects. Our objective is to develop assisting tool for clinicians.
All of the preprocessing methods are used to obtain good results. To show effectiveness of our proposed system, we have evaluated performance measures including accuracy, sensitivity, specificity, precision, and bar plot for the comparison of the proposed method with the existing systems. The paper is arranged as follows. Section 2 offers background knowledge on materials and methods. In Section 3, the experiments, results, and discussion are presented. Finally, conclusion and plan for future studies are presented in Section 4.
2. Materials and Methods
2.1. Materials
2.1.1. Dataset
In our study, the dataset is accessed from Open Access Series of Imaging Studies (OASIS). OASIS is a project, for compiling and sharing MRI datasets of the brain to make such data accessible to the scientific community. The data are accessible at http://www.oasisbrains.org. A sample of the MR brain image is shown in Figure 1.
(a) Normal
(b) Alzheimer’s disease
OASIS provides two types of data: crosssectional and longitudinal MRI data. In this study, we used crosssectional MRI data because we aimed to develop an automatic system for detecting AD, which would not require longitudinal data that had been gathered from AD patients over long periods of time.
The dataset consists of 416 subjects whose ages are between 18 and 96. In our study, we consider 126 samples (including 28 ADs and 98 HCs). Table 1 shows statistical information about the subjects included in the experiment. Only righthanded subjects are included in the study, consisting of both men and women. The exclusion criterion is patients less than 60 years of age or any of their reports are missing. The unbalanced data may cause difficulty in future recognition; we finetune the cost matrix to resolve this issue.

The dataset contains information about the patient’s demographics. The demographic features contain gender (M/F), age, education, socioeconomic status, and handedness. The mini mental state examination (MMSE) is a short 30point questionnaire test utilized to monitor for cognitive impairment and dementia. The MMSE test comprises simple questions and problems in numeral areas: the time and place, repeating list of words, arithmetic, language utilization, and comprehension, in addition to basic motor skills. Clinical dementia rating (CDR) is a numeric scale measuring the severity of symptoms of dementia. The patients’ cognitive and functional performances were accessed in six areas: memory, orientation, judgement and analytical, community affairs, residence and hobbies, and individual care. The patients’ CDR ranks and education level are listed in Tables 2 and 3, respectively.


2.2. Proposed Method
The proposed method consists of three important stages, namely, feature extraction using DTCWT, feature dimensionality reduction using PCA, and classification using feedforward artificial neural network. The overall block diagram of the suggested method is shown in Figure 2. Normalization of image is included in preprocessing section. All of these individual techniques have been proven outstanding, so we strongly believe that the proposed method can also achieve excellent results.
2.3. Image Preprocessing and Normalization
For each patient, each scanning session involves the MR of three or four T1weighted image scans. In order to add the signaltonoise ratio (SNR), all indicated MRI scans with the identical protocol of the same individual are motioncorrected and spatially coregistered, to the Talairach coordinate space to produce an averaged image, and then are brainmasked. The motion correction recorded the 3D images of all scans and then developed an average 3D image in initial acquisition space. Also, the scans are then resampled to 1 mm × 1 mm × 1 mm. The obtained image is converted from acquisition space to Talairach coordinate space. Lastly, the brain extraction is achieved.
We used MRIcro software (which can be downloaded from http://www.cabiatl.com/mricro/mricro/) and imported the image from the backup folder and then extracted the 2D MR image slices of each subject. In this paper, we only choose 32 important center slices from each subject manually based on our experience. These slices are used for preprocessing. The reason behind picking center slice from all slices is that it retains more relevant information about the brain tissues as compared to earlier slices and later slices in the group of 1–256 slices. The direction of the slice possibly may be sagittal, coronal, or axial. In this research, we chose axial direction by knowledge. The same process is applied to all the subjects (126 including both ADs and HCs). All images are in PNG format, and the dimensions of the slices are 176 × 208. The image is resized to 256 × 256 before being used for further processing.
2.4. Discrete Wavelet Transform
The discrete wavelet transform (DWT) is an image processing method [18] that gives multiscale representation of a stated signal or image [19]. Standard DWT is helpless to shift variance issue and only has horizontal and vertical directional selectivity [20]. Suppose denotes a particular signal, symbolizes the sampling point, and denote a highpass filter and lowpass filter, respectively, and and depict the coefficients of highpass and lowpass subbands. We have
The represents a lowpass filter along xaxis and highpass filter along yaxis. represents a highpass filter along xaxis and lowpass filter along yaxis. The represents lowpass filters along both directions, and represents highpass filters along both directions.
Here, the and have clearcut for both vertical and horizontal orientations. For the , it combines directions of both −45 and +45 degrees jointly, which stems from the utilization of realvalued filters in DWT. This combining also hinders the direction check [21].
2.5. DualTree Complex Wavelet Transform
The dualtree complex wavelet transform (DTCWT) is a modified version of the traditional DWT. To help boost the directional selectivity impaired by DWT, DTCWT is proposed. The traditional DWT is shift variant because of the decimation operation used in the transform. As a consequence, a small shift in the input signal can create a very dissimilar set of wavelet coefficients formed at the output. It utilizes two real DWTs processing input data in parallel [22]. The first DWT symbolizes the real component of the transform, whereas the second DWT depicts the imaginary component together forming a complex transform.
The DTCWT provides a solution for “shiftinvariant problems” as well as for “directional selectivity in two or more dimensions,” which are both shortcomings of the ordinary DWT [23]. It obtains directional selectivity by utilizing wavelets that are approximately analytic. It also has the ability to produce a total of six directionally discriminating subbands oriented in the ±15, ±45, and ±75 directions, for both the real (R) and imaginary (I) parts. Figure 3 illustrates the DTCWT. Let and be the filters in the first stage as in Figure 3. Let the new kth stage response of the first filter bank be and second filter bank be we now have the following result as a corollary of Lemma 1.
Corollary 1. Suppose one is provided with CQF pairs . For if and only if
A 2D image is decomposed by 2D DTCWT over a series of dilations and translations of a complicated scaling function and six complex wavelet functions that is, where gives the directionality of the complex wavelet function. This is to say that the decomposition of by utilizing the DTCWT creates one complex valued lowpass subband and six complex valued highpass subbands at every level of decomposition, where every highpass subband corresponds to one particular direction .
A study was carried out in [24] to compare the DTCWT’s directional selectivity to that of the DWT. The simulation results showed that the edges detected by the DTCWT had clear contours, and nearly all directions could be detected clearly and perfectly. However, the edges detected by the DWT were discontinuous, and only horizontal and vertical edges could be successfully detected. The results verified the effectiveness of the DTCWT over the DWT. By utilizing DTCWT, we are extracting DTCWT coefficients from the preprocessed images. The additional features include the information about the demographics of the patients such as age, gender, handedness, education, SES, and clinical examination. The handedness features are not included in the work since all the patients are righthandedness.
2.6. Principal Component Analysis
The coefficient from the DTCWT enlarges the dimensionality of feature space that makes the classification job more complicated.
Additionally, it leads to excessive computational overhead and enormous memory storage. As a result, it is essential to lower the dimension of the feature set and get the significant features to boost the classification result. Since the last two decades, a method called PCA has earned much more attention for data visualization and reduction of dimensionality. It systematically projects the initial input data to a lowerdimensional space, wellknown as principal subspace through an orthogonal transformation while preserving most of the data variations. For a stated set of likely correlated variables, these transformation outcomes in a set of values of linearly uncorrelated variables are called as principal components (PCs). All of the steps to implement PCA are demonstrated in Algorithm 1. The additional information on PCA and its implementations can be viewed in literature [25, 26].
Let us consider a set of data. PCA is employed to find a linear lowerdimensional reduction of the dataset. In this case, the variance of the constructed data is preserved. PCA limits the feature vectors to the component it selects, which leads to an effective classification algorithm. The main idea behind implementing PCA is reduction of the dimensionality of the DTCWT coefficients, which results in more adequate and accurate classification.
The following algorithm is utilized to obtain the principal components from the input matrix and finally fed to the feedforward neural network. Now, the input matrix possesses only these PCs. Hence, the size of the matrix is reduced. Therefore, feature extraction is done in two steps: DTCWT extracts the wavelet coefficients, and essential coefficients are later selected by the PCA as described in Algorithm 1.
 
Algorithm 1: PCA algorithm. 
2.7. FeedForward Neural Networks
2.7.1. Structure
Feedforward neural networks (FNN) are broadly used in pattern classification because they do not need any information regarding the probability distribution or a priori probabilities of distinct classes. Neural networks (NN) harness power from their densely parallel structure and their ability to acquire information from experience. As a result, they can be utilized for accurate classification of input data into different classes, provided that they are pretrained. The architecture of a multilayer feedforward neural network is shown in Figure 4.
Three factors need to be considered in designing an ANN for a specific application: (i) the topology of the network, (ii) the training algorithm, and (iii) the neuron activation function. A network may have many layers of neurons, and its complete architecture may possess either a feedforward or a back propagation structure. A multihiddenlayer back propagation NN with sigmoid neurons in its hidden layer is chosen. Similarly, linear neurons are selected for the output layer. The training vector is provided to the NN, which is instructed batch mode [27]. The NN is a twolayer network, and its configuration is represents the input neurons, is the hidden layer, and indicates that the brain under observation is either HC or AD.
2.7.2. Training Method
Mathematicians have already proven that a conjugate gradient (CG) algorithm, probing along conjugate gradient directions, produces a faster convergence than the steepest descent directions do. Among CG algorithm, the scaled conjugate gradient (SCG) method is the most powerful [28]. Thus, we utilize the SCG to train our network.
Let and be the connection weight matrix linking the input layer and hidden layer, and the hidden layer and the output layer, respectively. Later, we can deduce the training process reported by the following equations to improve these weighted values that can be divided into four subsequent steps [29]. (1)The calculation of the outputs of all neurons in the hidden layer is done by
Here, stands for the th input value, stands for the th output of the hidden layer, and refers to the activation function of hidden layer, commonly a sigmoid function as observed: (2)The outputs of all neurons in the output layer are stated as follows:
Here, represents the activation function of output layer that is usually a line function. At first, all weights are accredited with random values and amended by the delta rule on the basis to the learning samples. (3)The error is articulated as the MSE of the distinction among output and target value [30].where depicts the kth value of the genuine labels which is already wellknown to users and denotes the number of samples [31]. (4)Let us consider that there are samples; therefore, the fitness value can be written aswhere designates the vectorization of the (). The aim is to minimize the fitness function namely, force the output values of every sample appropriate to equivalent target values.
The hidden layer or the output layer is depicted in Figure 5. The inputs, weighted sum, and activation function of output layer are shown in Figure 5. The connection weight between the input layer and hidden layer and hidden layer and output layer is shown in Figure 6. The connection weights can also be represented in the matrix form known as connection weight matrix.
3. Experiment, Results, and Discussion
The proposed method is implemented using the 32bit Matlab 2015b environment on Intel(R) Core (TM) i32120, with a processing speed of 3.30 GHz and 2 GB of RAM running Microsoft Windows 7. Readers can repeat our results on any computer with which MATLAB is compatible.
This article aims at developing a CAD of AD brain system with better performance. The pseudocode is listed in Table 4.

3.1. Parameter Estimation for
It is always a major concern to find the optimum value of decomposition level . We know that a smaller s provides less information whereas a larger provides more information to the classifier. In order to avoid overfitting problem, a smaller is used. Here, we change the value of from 1 to 5 with increment of 1 and check up the corresponding average accuracies with FNN. The one which gives the highest accuracy is the optimal value of .
3.2. Feature Extraction
In this paper, we extract the DTCWT coefficients from the input images. The features of 5th resolution scales are selected because they provide higher classification performance than other resolution level scales. The DTCWT has a multiresolution representation as the wavelet transform does. For disease detection, it is preferable to use a few intermediate coefficient scales as the classifier input. The lowest scales have lost fine signal details whereas the most highly detailed scales contain mostly noise. Therefore, we prefer to choose only a few intermediate scales for the DTCWT coefficients. These obtained coefficients are sent as input to the PCA.
3.3. Feature Reduction
Excessive features increase calculation time as well as memory storage. In addition, they sometimes make classification much more complicated, which is known as curse of dimensionality. In this article, we utilized PCA to decrease the number of features.
Therefore, the extracted feature from DTCWT is sent to the PCA for the feature reduction. For each image, there are 768 features after 5th level of decomposition. As we have employed 32 slices for each patient, the total number of features becomes 32 × 768. Now, the image is reformed into a row vector of 1 × 24,576. The row vectors of 126 subjects are arranged into an “input matrix” with dimensions of 126 × 24,576. It is still too large for calculation. So, the input data matrix is now decomposed into the principal component “score matrix” and the “coefficient matrix.” The score matrix size after decomposition is 126 × 125. Here, the rows and columns of “score matrix” correspond to subjects and components, respectively.
The variance with the number of principal components from 1 to 18 is listed in Table 5. Experimenting with different numbers of principal components (PCs) revealed that accuracy with PC = 14 provided the best classification accuracy preserving 90.44% of the total variance. The curve of cumulative sum of variances with the number of principal component is shown in Figure 7. We did not set the energy threshold as 95% because that would cost too many features, along with computational burden.

3.4. BPNN Training
The 14 PCs are directly sent to BPNN. Thus, the number of input neurons is 14. Then, the number of hidden layer neurons () is determined as 10 according to the information entropy method [32]. Therefore, the architecture of the neural network becomes 14101. The SCG method is employed because it is extremely faster than BP, MBP, and ABP [28].
3.5. Performance Measures
There are several techniques to evaluate the efficiency of classifiers. The performance is calculated on the essence of the overall confusion matrix. It holds the correct and incorrect classification results. Table 6 shows a confusion matrix for binary classification, where , , , and depict true positive, true negative, false positive, and false negative, respectively, as illustrated in Table 7.


Here, AD brains are assumed to hold the value “true” and NC ones are assumed to hold the value “false” following normal convention.
The accuracy is the most accepted empirical measure to access effectiveness of classifier. It is formulated by
Sensitivity is the measure of the proportion of true positives that are correctly classified, and specificity is the measure of the proportion of negatives which are correctly classified. These are calculated by
The precision and the recall are formulated by
3.6. Statistical Analysis
In order to execute a strict statistical analysis, stratified crossvalidation (SCV) is used. We apply a 10fold CV technique in this experiment because of two reasons: (1) to make balance between reliable estimate and computational cost and (2) for providing a fair comparison because the common convention was to take the value of K equal to 10 [33].
A 10fold CV means we have to divide our dataset randomly into ten mutually exclusively folds of approximately equal size and almost the same distribution. In each run, 9 subsets will be used for training, and the remaining one will be utilized for the validation. This process is repeated 10 times, in which every subset is utilized for validation once. The 10fold CV is repeated 50 times; namely, a 50x 10fold CV is implemented.
The accuracies, sensitivities, and specificities obtained from the 50 runs of 10fold CV are presented in Table 8. Our method achieved an accuracy of 90.06 ± 0.01%, a sensitivity of 92.00 ± 0.04%, a specificity of 87.78 ± 0.04%, and a precision of 89.6 ± 0.03%.

3.7. Comparison to Other StateoftheArt Approaches
To further determine the effectiveness of the proposed “DTCWT + PCA + FNN,” we compared it with seven stateoftheart approaches in Table 8. Some of these approaches utilized different statistical settings, making direct comparison difficult. The results in Table 8 show that study [34–37] did not present standard deviations (SD) of three standards. The specificities of study [34–36] are lower than those demonstrated by other methods. Therefore, these three methods are not worthy for further study. Similarly, study [37] obtained a classification specificity of 100%. In spite of its high specificity, both the accuracy and sensitivity achieved by this algorithm are poor. Hence, this method is also not considered further for the study. Three other methods reported both mean values and standard deviation values. They also achieved satisfying results. Study [38] obtained promising results because of the voxelbased morphometry (VBM). Indeed, VBM has frequently been employed to study brain changes. Study [37] demonstrated that a taxi driver will normally have a larger back section of the posterior hippocampus. Study [39] concluded that global gray matter decreases linearly with old age but global white matter remains in the same amount. Nevertheless, VBM requires an accurate spatial normalization, or the classification accuracy may decrease significantly. Study [40] was based on a novel approach called the displacement field (DF). This study measured and estimated the displacement field of various slices between AD and HC subjects. There are other methods that have distinguished AD from HC; however, they dealt with images formed by other modalities: PET, SPECT, DTI, and so forth. Hence, they are also not considered in this study.
Finally, the proposed “DTCWT + PCA + FNN” achieved an accuracy of 90.06 ± 0.01%, a sensitivity of 92.00 ± 0.04%, a specificity of 87.78 ± 0.04%, and a precision of 89.60 ± 0.03%. With respect to classification accuracy, our approach outperforms five other methods and is almost equal to the accuracies of the remaining two methods that did not account for means and standard deviations. We also achieved a promising sensitivity and a promising specificity. Hence, our results are either better than or comparable to those of the other methods. The bar plot of the algorithm comparison is shown in Figure 8. The acronyms list is depicted in Table 9.

4. Conclusions and Future Research
We presented an automated and accurate method for AD identification based on a DTCWT, PCA, and FNN. The results showed that the proposed method achieved an accuracy of 90.06 ± 0.01%, a sensitivity of 92.00 ± 0.04%, a specificity of 87.78 ± 0.04%, and a precision of 89.6 ± 0.03% and outperformed 7 stateoftheart algorithms.
We will focus our future research on the following aspects: (i) testing other advanced variants of wavelet such as 3DDTCWT, wavelet packet analysis, and fractional calculus; (ii) utilizing different feature reduction techniques such as independent component analysis (ICA) [41], linear discriminant analysis (LDA) [42], probabilistic PCA [43], or sparseautoencoders [44]; (iii) testing our data with leastsquare techniques [45], kernel support vector machine (kSVM), such as fuzzy SVM [46], radial basis function neural network (RBFNN) [47], deep learning methods such as convolutional neural network (CNN) [48], and other alternative pattern recognition tool for classification; (iv) utilizing advanced swarm intelligence techniques such as particle swarm optimization [49], artificial bee colony [50], genetic pattern search [51], ant colony optimization [52], and biogeographybased optimization [53] to find the optimal kernel; (v) testing the proposed method on images obtained from different modalities such as computed tomography (CT) [54], ultrasound, spectrum imaging [55], and 3D MRI; (vi) utilizing other advance image preprocessing technique to enhance the classification performance, such as image denoising, image enhancement, and image segmentation; and (vii) classification may be carried out on the sparsity domain.
Conflicts of Interest
The authors declare that they have no conflict of interest.
Acknowledgments
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 (NRF2014M3C7A1046050). And this study was financed by the research fund of Chosun University, 2017.
References
 R. Brookmeyera, E. Johnsona, K. ZieglerGrahamb, and H. M. Arrighic, “Forecasting the global burden of Alzheimer’s disease,” Alzheimer’s & Dementia, vol. 3, no. 3, pp. 186–191, 2007. View at: Publisher Site  Google Scholar
 C. Davatzikos, P. Bhatt, L. M. Shaw, K. N. Batmanghelich, and J. Q. Trojanowski, “Prediction of MCI to AD conversion, via MRI, CSF biomarkers, and pattern classification,” Neurobiology of Aging, vol. 32, no. 12, pp. 2322.e19–2322.e27, 2011. View at: Publisher Site  Google Scholar
 A. Nordberg, J. O. Rinne, A. Kadir, and B. Langstrom, “The use of PET in Alzheimer disease,” Nature Reviews Neurology, vol. 6, no. 2, pp. 78–87, 2010. View at: Publisher Site  Google Scholar
 M. D. Greicius, G. Srivastava, A. L. Reiss, and V. Menon, “Defaultmode network activity distinguishes Alzheimer’s disease from healthy aging: evidence from functional MRI,” Proceedings of the National Academy of Sciences of the United States of America, vol. 101, no. 13, pp. 4637–4642, 2004. View at: Publisher Site  Google Scholar
 T. Magnander, E. Wikberg, J. Svensson et al., “A novel statistical analysis method to improve the detection of hepatic foci of 111 Inoctreotide in SPECT/CT imaging,” EJNMMI Physics, vol. 3, no. 1, p. 1, 2016. View at: Publisher Site  Google Scholar
 P. De Visschere, M. Nezzo, E. Pattyn, V. Fonteyne, C. Van Praet, and G. Villeirs, “Prostate magnetic resonance spectroscopic imaging at 1.5 tesla with endorectal coil versus 3.0 tesla without endorectal coil: comparison of spectral quality,” Clinical Imaging, vol. 39, no. 4, pp. 636–641, 2015. View at: Publisher Site  Google Scholar
 E. D. Roberson and L. Mucke, “100 years and counting: prospects for defeating Alzheimer’s disease,” Science, vol. 314, no. 5800, pp. 781–784, 2006. View at: Publisher Site  Google Scholar
 M. Tabaton, P. Odetti, S. Cammarata et al., “Artificial neural networks identify the predictive values of risk factors on the conversion of amnestic mild cognitive Impairment,” Journal of Alzheimer’s Disease, vol. 19, no. 3, pp. 1035–1040, 2010. View at: Publisher Site  Google Scholar
 B. S. Mahanand, S. Suresh, N. Sundararajan, and K. M. Aswatha, “Identification of brain regions responsible for Alzheimer’s disease using selfadaptive resource allocation network,” Neural Networks, vol. 32, pp. 313–322, 2012. View at: Publisher Site  Google Scholar
 B. Jeurissen, A. Leemans, and J. SIjbers, “Automated correction of improperly rotated diffusion gradient orientations in diffusion weighted MRI,” Medical Image Analysis, vol. 18, no. 7, pp. 953–962, 2014. View at: Publisher Site  Google Scholar
 L. Hamelin, M. Bertoux, M. Bottlaender et al., “Sulcal morphology as a new imaging marker for the diagnosis of early onset Alzheimer’s disease,” Neurobiology of Aging, vol. 36, no. 11, pp. 2932–2939, 2015. View at: Publisher Site  Google Scholar
 M. Á. Caballero, M. Brendel, A. Delker et al., “Mapping 3year changes in gray matter and metabolism in Abetapositive nondemented subjects,” Neurobiology of Aging, vol. 36, no. 11, pp. 2913–2924, 2015. View at: Publisher Site  Google Scholar
 G. Yang, Y. Zhang, J. Yang et al., “Automated classification of brain images using wavelet energy and biogeographybased optimization,” Multimedia Tools and Applications, vol. 75, no. 23, pp. 15601–15617, 2016. View at: Publisher Site  Google Scholar
 E. S. ElDahshan, H. M. Mohsen, K. Revett, and A. B. Salem, “Computeraided diagnosis of human brain tumor through MRI: a survey and a new algorithm,” Expert Systems with Applications, vol. 41, no. 11, pp. 5526–5545, 2014. View at: Publisher Site  Google Scholar
 Y. Zhang, S. Wang, G. Ji, and Z. Dong, “An MR images classifier system via particle swarm optimization and kernel support vector machine,” The Scientific World Journal, vol. 2013, Article ID 130134, 9 pages, 2013. View at: Publisher Site  Google Scholar
 I. Álvarez, J. M. Górriz, J. Ramírez et al., “Alzheimer’s diagnosis using eigenbrains and support vector machines,” Electronics Letters, vol. 45, no. 7, pp. 973–980, 2009. View at: Publisher Site  Google Scholar
 M. Z. Iqbal, A. Ghafoor, A. M. Siddhiqui, M. M. Riaz, and U. Khalid, “Dualtree complex wavelet transform and SVD based medical image resolution enhancement,” Signal Processing, vol. 105, pp. 430–437, 2014. View at: Publisher Site  Google Scholar
 Y. Zhang, B. Peng, S. Wang et al., “Image processing methods to elucidate spatial characteristics of retinal microglia after optic nerve transection,” Scientific Reports, vol. 6, article 21816, 2016. View at: Publisher Site  Google Scholar
 D. K. Shin and Y. S. Moon, “Superresolution image reconstruction using wavelet based patch and discrete wavelet transform,” Journal of Signal Processing Systems, vol. 81, no. 1, pp. 71–81, 2015. View at: Publisher Site  Google Scholar
 Y. Zhang, S. Wang, G. Ji, and Z. Dong, “Exponential wavelet iterative shrinkage thresholding algorithm with random shift for compressed sensing magnetic resonance imaging,” IEEJ Transcations on Electrical and Electronic Engineering, vol. 10, no. 1, pp. 116117, 2015. View at: Publisher Site  Google Scholar
 S. Beura, B. Majhi, and R. Dash, “Mammogram classification using two dimensional discrete wavelet transform and graylevel cooccurance matrix for detection of breast cancer,” Neurocomputing, vol. 154, pp. 1–14, 2015. View at: Publisher Site  Google Scholar
 N. Kingsbury, “Complex wavelets for shift invariant analysis and filtering of signals,” Applied and Computational Harmonic Analysis, vol. 10, no. 3, pp. 234–353, 2001. View at: Publisher Site  Google Scholar
 A. Barri, A. Dooms, and P. Schelkens, “The near shiftinvariance of the dualtree complex wavelet transform revisited,” Journal of Mathematical Analysis and Applications, vol. 389, no. 2, pp. 1303–1314, 2012. View at: Publisher Site  Google Scholar
 S. Wang, S. Lu, Z. Dong, J. Yang, M. Yang, and Y. Zhang, “Dualtree wavelet transform and twin support vector machine for pathological brain detection,” Applied Science, vol. 6, no. 6, p. 169, 2016. View at: Publisher Site  Google Scholar
 C. M. Bishop, Pattern Recognition and Machine Learning, Springer, New York, 2006.
 R. O. Duda, P. E. Hart, and D. G. Stork, Pattern Classification, John Wiley & Sons, 2012.
 D. Guo, Y. Zhang, Q. Xiang, and Z. Li, “Improved radio frequency identification method via radial basis function neural network,” Mathematical Problems in Engineering, vol. 2014, Article ID 420482, 9 pages, 2014. View at: Publisher Site  Google Scholar
 Y. Zhang, Z. Dong, L. Wu, and S. Wang, “A hybrid method for MRI brain image classification,” Expert Systems with Applications, vol. 38, no. 8, pp. 10049–10053, 2011. View at: Publisher Site  Google Scholar
 M. Manoochehri and F. Kolahan, “Integration of artificial neural network and simulated annealing algorithm to optimize deep drawing process,” The International Journal of Advaned Manufacturing Technology, vol. 73, no. 1, pp. 241–249, 2014. View at: Google Scholar
 S. U. Aswathy, G. G. D. Dhas, and S. S. Kumar, “A survey on detection of brain tumor from MRI brain images,” in 2014 International Conference on Control, Instrumentation, Communication and Computational Technologies (ICCICCT), pp. 871–877, Kanyakumari, 2014. View at: Publisher Site  Google Scholar
 A. Poursamad, “Adaptive feedback linearization control of anticlock braking systems using neural networks,” Mechatronics, vol. 19, no. 5, pp. 767–773, 2009. View at: Publisher Site  Google Scholar
 H. C. Yuan, F. L. Xiong, and X. Y. Huai, “A method for estimating the number of hidden neurons in feedforward neural networks based on information entropy,” Computers and Electronics Agriculture, vol. 40, no. 1–3, pp. 57–64, 2003. View at: Publisher Site  Google Scholar
 Y. Zhang, S. Wang, P. Phillips, and G. Ji, “Binary PSO with mutation operator for feature selection using decision tree applied to spam detection,” KnowledgeBased Systems, vol. 64, pp. 21–31, 2014. View at: Google Scholar
 C. Plant, S. J. Teipel, A. Oswald et al., “Automated detection of brain atrophy patterns based on MRI for the prediction of Alzheimer’s disease,” NeuroImage, vol. 50, no. 1, pp. 162–174, 2010. View at: Publisher Site  Google Scholar
 D. Jha and G. R. Kwon, “Alzheimer disease detection in MRI using curvelet transform with KNN,” The Journal of Korean Institute of Information Technology, vol. 14, no. 8, 2016. View at: Publisher Site  Google Scholar
 Y. Zhang, S. Wang, and Z. Dong, “Classification of Alzheimer disease based on structural magnetic resonance imaging by kernel support vector machine decision tree,” Progress in Electromagnetic Research, vol. 144, pp. 171–184, 2014. View at: Publisher Site  Google Scholar
 E. A. Maguire, D. G. Gadian, I. S. Johnsrude et al., “Navigationrelated structural change in the hippocampi of taxi drivers,” Proceedings of the National Academy of Sciences, vol. 97, no. 8, pp. 4398–4403, 2000. View at: Publisher Site  Google Scholar
 K. R. Gray, P. Alijabar, R. A. Heckemann, A. Hammers, and D. Rueckert, “Random forestbased similarity measures for multimodal classification of Alzheimer’s disease,” NeuroImage, vol. 65, pp. 167–175, 2013. View at: Publisher Site  Google Scholar
 C. D. Good, I. S. Johnsrude, J. Ashburner, R. N. A. Henson, K. J. Friston, and R. S. J. Frackowiak, “A voxelbased morphometric study of ageing in 465 normal adult human brains,” NeuroImage, vol. 14, no. 1, Part 1, pp. 21–36, 2001. View at: Publisher Site  Google Scholar
 Y. Zhang and S. Wang, “Detection of Alzheimer’s disease by displacement field and machine learning,” PeerJ, vol. 3, article e1251, 2015. View at: Publisher Site  Google Scholar
 S. Makeig, T.P. Jung, A. J. Bell, D. Ghahremani, and T. Sejnowski, “Blind separation of auditory eventrelated brain responses into independent components,” Proceedings of the National Academy of Sciences of the United States of America, vol. 94, no. 20, pp. 10979–10984, 1997. View at: Publisher Site  Google Scholar
 A. J. Izenman, “Linear discriminant analysis,” in Modern Multivariate Statistical Techniques, pp. 237–280, Springer, New York, 2013. View at: Google Scholar
 M. E. Tipping and C. M. Bishop, “Probabilistic principal component analysis,” Journal of the Royal Statistical Society: Series B (Statistical Methodology), vol. 61, no. 3, pp. 611–622, 1999. View at: Publisher Site  Google Scholar
 B. A. Olshausen and D. J. Field, “Sparse coding with an overcomplete basis set: a strategy employed by V1,” Vision Research, vol. 37, no. 23, pp. 3311–3325, 1997. View at: Publisher Site  Google Scholar
 K. Polat, S. Güneş, and A. Arslan, “A cascade learning system for classification of diabetes disease: generalized discriminant analysis and least square support vector machine,” Expert Systems with Applications, vol. 34, no. 1, pp. 482–487, 2008. View at: Publisher Site  Google Scholar
 S. Wang, X. Yang, Y. Zhang, P. Phillips, J. Yang, and T.F. Yuan, “Identification of green, oolong and black teas in china via wavelet packet entropy and fuzzy support vector machine,” Entropy, vol. 17, no. 10, pp. 6663–6682, 2015. View at: Publisher Site  Google Scholar
 M. J. Er, S. Wu, J. Lu, and H. L. Toh, “Face recognition with radial basis function (RBF) neural networks,” IEEE Transactions on Neural Networks, vol. 13, no. 3, pp. 697–710, 2002. View at: Publisher Site  Google Scholar
 A. Krizhevsky, I. Sutskever, and G. E. Hinton, “Imagenet classification with deep convolutional neural networks,” Advances in Neural Information Processing Systems, pp. 1097–1105, 2012. View at: Google Scholar
 Y. Zhang, S. Wang, P. Phillips, and G. Ji, “Binary PSO with mutation operator for feature selection using decision tree applied to spam detection,” KnowledgeBased Systems, vol. 64, pp. 22–31, 2014. View at: Publisher Site  Google Scholar
 Y. Zhang, L. Wu, and S. Wang, “Magnetic resonance brain image classification by an improved artificial bee colony algorithm,” Progress in Electromagnetics Research, vol. 116, pp. 65–79, 2011. View at: Publisher Site  Google Scholar
 Y. Zhang, S. Wang, G. Ji, and Z. Dong, “Genetic pattern search and its application to brain image classification,” Mathematical Problems in Engineering, vol. 2013, Article ID 580876, 8 pages, 2013. View at: Publisher Site  Google Scholar
 C. Pang, G. Jiang, S. Wang et al., “Gene order computation using Alzheimer’s DNA microarray gene expression data and the ant colony optimisation algorithm,” International Journal of Data Mining and Bioinformatics, vol. 6, no. 6, pp. 617–632, 2012. View at: Publisher Site  Google Scholar
 S. Wang, Y. Zhang, G.L. Ji, J.G. Wu, and L. Wei, “Fruit classification by waveletentropy and feedforward neural network trained by fitnessscaled chaotic ABC and biogeographybased optimization,” Entrophy, vol. 17, no. 8, pp. 5711–5728, 2015. View at: Publisher Site  Google Scholar
 Y. Chen, Y. Zhang, J. Yang et al., “Curvelike structure extraction using minimal path propagation with backtracking,” IEEE Transactions on Image Processing, vol. 25, no. 2, pp. 988–1003, 2016. View at: Publisher Site  Google Scholar
 Z. Dong, Y. Zhang, F. Liu, Y. Duan, A. Kangarlu, and B. S. Peterson, “Improving the spectral resolution and spectral fitting of H MRSI data from human calf muscle by the SPREAD technique,” NMR in Biomedicine, vol. 27, no. 11, pp. 1325–1332, 2014. View at: Publisher Site  Google Scholar
 Y. Zhang, Z. Dong, P. Phillips et al., “Detection of subjects and brain regions related to Alzheimer’s disease using 3D MRI scans based on eigenbrain and machine learning,” Frontlier in Computational Neuroscience, vol. 9, p. 66, 2015. View at: Publisher Site  Google Scholar
Copyright
Copyright © 2017 Debesh Jha et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.