(IJACSA) International Journal of Advanced Computer Science and Applications,

Vol. 14, No. 1, 2023

Modeling of Whale Optimization with Deep Learning

based Brain Disorder Detection and Classification
Uvaneshwari M1, Dr. M. Baskar2*
Research Scholar-Department of Computer Science and Engineering-School of Computing-College of Engineering and
Technology, SRM Institute of Science and Technology, Kattankulathur, Chengalpattu, Tamilnadu, 603 203, India1
Associate Professor-Department of Computing Technologies-School of Computing-College of Engineering and Technology,
SRM Institute of Science and Technology, Kattankulathur, Chengalpattu, Tamilnadu, 603 203, India2

Abstract—Brain disorders are a significant source of Processing steps or appropriate architectural design must be
economic strain and unfathomable suffering in modern society. defined in advance [5]. Generally, a classification study using
Imaging techniques help diagnose, monitor and treat mental the ML method needs four stages: feature selection, feature
health, neurological, and developmental disorders. To aid in the extraction, feature-based classification algorithm selection,
Computer-Aided Diagnosis of brain diseases, deep learning (DL) and dimensionality reduction [6]. This process requires
was used for the analysis of neuroimages from modalities multiple optimization steps and professional knowledge that
including Positron Emission Tomography (PET), Structural might take considerable time [7]. To conquer this problem,
Magnetic Resonance Imaging (SMRI), and functional MRI. In deep learning (DL), an emerging field of ML study that uses
this study, a Whale Optimization Algorithm is used with Deep
raw neuroimaging information to generate features through
Learning to analyse MRI scans for signs of neurological disease.
WOADL-BDDC may detect and label abnormalities in the brain
"on-the-fly" learning, is received significant interest in the
based on an MRI scan. It uses a two-step pre-processing Area of large-scale, higher-dimension medical imaging
procedure, first using guided filtering to get rid of background analysis. DL approaches, like convolution neural networks
noise and then using U-Net segmentation to get rid of the top of (CNN), have proved to outperform the present ML method [8,
the head. QuickNAT, along with RMSProp, is used to segment 9]. Compared with conventional ML techniques, the DL
the brain. When analysing data, WOADL-BDDC uses radionics approach automatically discovers the proper representation
to collect information from every layer. When used in a without specialized knowledge and permits the non-expert to
convolutional recurrent neural network model, the Whale utilize DL approaches efficiently. Thus, the DL method has
Optimization Algorithm may accurately categorize mental rapidly become a method of choice for medical image analysis
illness. WOADL-BDDC is put through its paces using ADNI 3D. [10].
Compared to state-of-the-art classification results from Vgg16,
Graph CNN, Modified ResNet18, Non-linear SVM, ResNet50- This study progresses a Whale Optimization Algorithm
SVM, ResNet50-RF, the suggested technique achieved the with Deep Learning based Brain Disorder Detection and
greatest accuracy. It was demonstrated that the suggested model Classification (WOADL-BDDC) on MRI images. The
is superior to other models for classification from MRI images. WOADL-BDDC technique initially performs pre-processing
In simulations, the proposed approach is shown to be effective in in 2 stages: guided filtering (GF) based noise elimination and
optimizing hyperparameters with an accuracy of 94.38 % on TR U-Net segmentation-based skull stripping. Besides, the
set and 94.87% on TS set, Precision of 96.43% on TR set and QuickNAT with RMSProp optimizer is applied to segment the
97.62% on TS set, and an F1-Score of 89.35 % and 92.10% on brain parts. For feature extraction, the WOADL-BDDC
TR and TS set, respectively. technique performs a radionics approach that separates
features of all aspects. Finally, the WOA with convolutional
Keywords—Brain disorder detection; magnetic resonance recurrent neural network (CRNN) model is applied to classify
imaging; deep learning; convolutional recurrent neural network; brain disorders. The experimental evaluation of the WOADL-
whale optimization algorithm BDDC technique is tested on the ADNI 3D dataset.
I. INTRODUCTION The wide range in tumour size, shape, and location poses a
The brain is the management center of the central nervous significant problem for diagnosing brain tumours. The
system and is liable for implementing activities all over the scientific community can benefit from this review since it
human body [1, 2]. As in other medical sectors, there exist provides a thorough literature on brain tumour detection using
Innovative neuroimaging technologies comprising positron magnetic resonance imaging. The healthcare sector is unlike
emission tomography (PET), and MRI was introduced and any other. Regardless of financial constraints, this is a top
utilized for identifying Alzheimer's Disease (AD) based priority area that attracts customers expecting only the best
molecular and structural biomarkers [3,17]. Rapid progression service. Given deep learning's track record of success in the
in neuroimaging techniques makes them challenging to real world, it should be no surprise that this promising
incorporate largescale, higher-dimension multimodal technique is already playing a crucial role in medical imaging,
neuroimaging datasets [4]. Thus, computer-aided machine yielding promising results with high accuracy. The brain is the
learning (ML) algorithms for integrative analysis have gained master organ, regulating the functions of every other body
considerable interest. To apply this ML approach,Pre- component.
*Corresponding Author.

298 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 1, 2023

Due to the intricacy of size and position variations, Brain tumours are notoriously difficult to diagnose
automated brain tumour recognition in MRI is challenging. because of the wide range of tumour sizes, forms, and shapes
The pictures produced from MRI scans of the brain are that must be considered during the process [25]. Surgical,
processed using proposed statistical, morphological, and radiological, and chemotherapeutic options are all on the table
thresholding methods to detect tumours.Based on our findings, when a brain tumour diagnosis has been made, and prognoses
we can confidently state that conventional algorithms perform can be formulated based on the tumor's nature and location
admirably for both the number of initial clusters and the [26]. Therefore, early diagnosis and identification of brain
locations of their centers. Pixel classification becomes difficult tumours aid in treatment planning and patient monitoring. It's
if these groups shift depending on the parameters you use to a vital factor in better treatment and increased survival rates.
construct them. As it is, the cluster centroid value is chosen at To learn more about tumours, many medical imaging and
random in the most common fuzzy cluster mean procedure. diagnostic methods are performed. It is possible to distinguish
Because of this, it will take longer to get the intended result. between usually and abnormally expanding brain cells with
Radiologists have to spend a lot of time manually segmenting the aid of imaging techniques like magnetic resonance
and evaluating MRI brain pictures. The segmentation is imaging (MRI) and computed tomography (CT)[27].
performed using machine learning approaches that are less Computed Tomography (CT) scans are utilized for diagnosis
accurate and have a slower calculation speed. Tumor through X-ray and computer to produce pictures of the
categorization and identification using neural network patient's brain in axial slices [28].
methods have been attempted extensively but with mixed
results. Detection accuracy relies on the segmentation and According to the systematic literature evaluation, there is
detection algorithms employed. Thus far, current systems need space for improvement in brain tumour detection. Because
to provide better picture quality and precision. brain tumours can be any size or shape, current segmentation
methods need further work before they can be used for
A convolutional recurrent neural network (CRNN) will be tumours. The importance of enhancement and segmentation in
employed to categorize the tumors in the picture. Increased tumour identification stems from their role in addressing the
precision and decreased repetition are the only two benefits of shortcomings of current approaches.
this method.
The significance of medical data offered along with
II. LITERATURE REVIEW imaging data can be emphasized by incorporating clinical
features, including texture-related features, for generating
The authors [11] modeled a deep ensemble learning hybrid feature vectors. Though several computer-aided
structure to harness DL techniques to tap the 'wisdom of diagnosis models are existed in the literature to perform the
experts' and compile multisource data. At the voting layer, two brain tumor classification processes, it is required to improve
sparse AEs were instructed for feature learning to minimize the classifier results. Due to the repeated extension of the
the correlation of features and expand the base classifier model, the number of parameters of DL models also rises
eventually. At the stacking layer, a non-linear feature- rapidly, which leads to model overfitting. So, the WOA
weighted technique related to a DBN was devised to rank the algorithm is used to select CRNN parameters. Personalized
base classifiers, which disrupt the conditional independence. image collection based on estimates has been proven to be
In [12], the authors intend to formulate an innovative DL capable of seeing channels during a CT examination, thanks to
technique for detecting or productively predicting AD. The the ongoing development of AI. Significant learning, in
authors devised densely connected CNNs with a connection- particular, has demonstrated encouraging outcomes in
wise attention system for learning the multilevel attributes of automated planning. It's so good at processing visual
brain MRI for AD categorization. The authors employed the information that we can now employ neural connections to
densely associated NN for deriving multiscale features in pre- construct meaningful images rather than merely sort them. In
processing images. A connection-wise attention system has particular, CRNN Association is well suited for evaluating
been implemented for integrating associations among images from channels like CT and X-ray. CRNNs are mainly
attributes from distinct layers. In [13], the authors employ a designed to handle photographs better and finish picture
DL-related Granger causality estimation of brain connectivity groups. In this way, CRNNs can approach radiologists'
structure. It uses the robustness of LSTM in changing time accuracy when recognizing crucial elements in CT channels or
series processing. The authors employ MRI to examine the other signature images.
cerebral cortex property and exploit rs-fMRI to analyze the
function network's graph metric. In [14], the authors establish III. THE PROPOSED WOADL-BDDC MODEL
that it is settled well with the multi-task learning method. In
[15] devised a DL technique for every level of feature In this study, we have established a new WOADL-BDDC
extracting process and fuzzy hyper-plane related least square technique for recognizing brain disorders on MRI images. The
twin SVM (FLS-TWSVM) for classifying the derived features WOADL-BDDC approach encompasses GF-based noise
for initial analysis of AD through removing sagittal plane removal, U-Net-based skull stripping, QuickNAT with
slices from 3-D MRI imageries. In [16], an Alzheimer's RMSProp-based brain segmentation, radionics feature
classification and detection method was provided. The BoVW extraction, CRNN classification, and WOA hyperparameter
method was employed to enhance the efficacy of texture- tuning. Fig. 1 represents the block diagram of the WOADL-
related features, like the histogram of gradient, GLCM, and BDDC system.
scale-invariant feature transform, LBP.

299 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 1, 2023

A. Image Pre-Processing
A GF is an edge-preserving smoothing light filter. It
applies edge-preserving smoothing on the image, utilizing the
content of the secondary image, named a guidance image, to
impact the filtering. The guidance image could be a
completely different image or different versions of an image.
The bilateral filter might filter out texture or noise when
maintaining sharp edges. A Gaussian operator can compute
the convolution operator, and a proposal of Gaussian
smoothening can be made by convolution. The Gaussian
operator can be defined using the following equations:
( )
( ) (1)
√

The optimum smoothening filter for the image can be

localized in the frequency and spatial domains, where the
ambiguity relativity can be satisfied as follows [19]:
(2)
The Gaussian operator in 2D can be represented by:
( )
( ) (3)
Where (sigma) signifies the standard deviation (SD) of Fig. 1. Block diagram of WOADL-BDDC system.
Gaussian and( ) represents the Cartesian coordinates of the
image. In medical applications, it serves as an initial phase In contrast, it compensates for higher-class imbalanced
since it identifies path abnormality in the speed, brain, and datasets and encourages the accurate assessment of anatomical
accuracy of diagnoses. It isolates non-brain tissues such as the boundaries using a proper weighting system. The two primary
scalp and skull from the brain image. Skull stripping removes techniques of QuickNAT are the training method with
unwanted components and non-brain anatomy from scanned auxiliary labels and the FCNN framework. The trained model
images. There exist five convolution blocks in the and code are presented as extensions of MatConvNet for off-
downsampling route. Currently, there are overall of 1024 the-shelf usage.
feature maps. Excepting the last block, max pooling with C. Radiomics Feature Extraction
stride 22 can be performed after the completion of every block
for downsampling. The feature map is reduced in size from For the process of extracting features, the presented
240x240 to 1515. Each up-sampling block initiates by a algorithm will use radionics features. The abovementioned
deconvolution layer of stride 22 and filters size 33. features are derived by in-house software, utilizing
Consequently, feature map is becoming increasingly popular. PyRadiomics and python's skit-learn package [19]. Texture,
The two convolution layers in the up-sampling block lower intensity, and shape are the three kinds of features that are
the quantity of deconvolution feature map and feature map in calculated. Twenty-six shape-based and intensity-based
the encoding route. features can be extracted for all extraction settings (width,
filter, bin, and contour). With sixteen distinct stages, which
B. Brain Segmentation can be permutations of 2 contours (HGG and Gold),
In this work, the QuickNAT carries out the segmentation bandwidth (2, 4, 8, 16), and two filters (LoG and original) and
of brain regions with RMSPropoptimizer.QuickNAT is the the entire count of derived radionics imaging features was
first method that uses an unlabelled neuroimaging dataset with 1450.
auxiliary labels for training and a higher testing speed of the D. CRNN-Based Brain Disorder Classification
FC network for brain segmentation [18]. QuickNAT
comprises sagittal, coronal, and axial views monitored by a The CRNN model is applied here to detect and classify
view aggregation stage to deduce the last segment. Every brain disorders. The study presents the CRNN utilized for
FCNN is a similar structure and is stimulated by the organizing MRI [20]. Initially, CNN is used for learning
conventional encoding/decoding-based UNet structure with higher-level feature representation. Next, the CNN-learned
skip connection improved with an un-pooling layer. Also, the feature was fed into the Bi-GRU layer for learning the
study presents dense links in all the encoding/decoding blocks temporal correlation data. At last, this feature was provided in
to promote feature re-usability and to help gradient flow, that the FC layer by function for the outcome of the
is, the small number of training datasets. A multi-class Dice likelihood distribution of distinct classes. In the presented
loss and weighted logistic loss can enhance the network. method, the convolution RNN has encompassed two BiGRU
layers ( ) and eight convolution layers ( ). The
network parameter and architecture are given below:

300 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 1, 2023

: The first two stacked convolution layer uses 32 Whereas represents the best position of the attained
filters with the stride of (1,1) and receptive field of (3,5). solution, characterizes the count of iterations, and the
Then, -pooling with the stride of (4,3) decreases the size constant variables and are characterized by:
of feature maps.
: The subsequent two convolution layers use 64 (6)
filters with a stride of (1,1) and a receptive field of (3,1) and (7)
are utilized for learning local patterns alongside the frequency
dimension. It can be -pooling with the stride of (4,1). Now, denotes an arbitrary vector amongst [0, 1], and
variable is an arbitrary vector utilized for controlling the
: The succeeding pairs of convolution layer applied convergence procedure. It can be linearly decreased from two
128 filters with a stride of (1,1) and receptive field of (1,5) and to zero through iteration:
are utilized for learning local patterns along the time
dimension. It can subsequently be the -pooling with the (8)
stride of (1,3).
In Eq. (5), indicates the present iteration, and shows
: The successive two convolution layer uses 256
filters with a stride of (1,1) and receptive field of (3,3) for the maximal iteration count of the vector utilized for
learning joint time–frequency features. It can subsequently be displaying the transition method amongst the exploratory and
the -pooling of (2,2) stride. exploitative actions.

: 2BiGRU layers with 256 cells were utilized for In WOA, the exploitation stage could be one of 2 models,
temporal summarization, and he activation function was used. shrinking surroundings or spiral‐shaped methods. During the
The dropout with the possibility of 0:5 is applied to every Bi- shrinking surrounding model, it is accomplished by reducing
GRU layer to prevent over-fitting. Fig. 2 illustrates the the value. The spiral path model upgrades the distance betwixt
framework of CRNN. the most efficient search agents attained up to this point ( )
and searching agents (X), then boosts a novel location for the
E. Hyperparameter Tuning searching agent based on the subsequent as:
For adjusting the hyperparameters related to the CRNN ( ) ( ) ( ) (9)
model, the WOA is utilized in this work. WOA is a
population-based heuristic approach that models the hunting The probabilities value of 0.5 is set as follows:
strategy of humpback whales [21]. They construct unique
maneuvers for paralyzing the prey by swimming nearby them ( ) ( )
( ) { (10)
in a spiral or circle-shaped path. This technique is named the ( )
bubble-net attack. The hunting behavior can be specified into
two stages. The initial stage is the exploration phase, where Whereas denotes an arbitrary probability value. The
the whales randomly seek prey. The next stage is the pseudocode of the proposed model is shown in Algorithm 1.
exploitation stage, whereby the whales implement bubble-net Algorithm 1: Overall process of WOADL-BDDC technique
hunting. The following equation is utilized for mathematically
modeling these behaviors as follows: Input: ADNI 3D Dataset
Output: Classified images
⃗ | ( ) ( )| (4) Start
( ) ( ) (5) For every image
do
Image_Pre-processing (GF_noise, U_Net skull stripping)
Segment_image()
QuickNAT mode
Hyperparameter adjustment using RMSProp Optimizer
End
Feature_extraction(radiomics)
Determine texture, shape, intensity
Classify()
Execute CRNN
Hyperparameter Tuning using WOA
Execute Exploration stage
Execute Exploitation stage
End
Fig. 2. Framework of CRNN. End

301 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 1, 2023

IV. RESULTS AND DISCUSSION Fig. 4 determines the classifier results of the WOADL-
The presented WOADL-BDDC system is simulated using BDDC system under the training (TR) set and testing (TS) set.
a Python tool. The results are examined under the ADNI 3D Fig. 4(a) represents the confusion matrix offered by the
dataset [22]. WOADL-BDDC model under the TR set. The figure
symbolized that the WOADL-BDDC model had identified 28,
51, and 5 instances under CN, MCI, and AD classes. Also,
Fig. 4(b) describes the confusion matrix offered by the
WOADL-BDDC methodology under the TS set. The figure
exemplified that the WOADL-BDDC algorithm has identified
4, 26, and 7 instances under CN, MCI, and AD classes.
Fig. 4(c) and 4(d) validate the precision-recall analysis of the
WOADL-BDDC system under training and TS sets. The
figures described that the WOADL-BDDC method had
acquired maximal precision-recall performance under all
varying classes.
Table I reports the overall brain disorder classification
outcomes of the WOADL-BDDC model. Fig. 5 represents a
brief brain disorder classifier results of the WOADL-BDDC
model under the TR set. The results indicated that the
WOADL-BDDC model had shown enhanced results under all
measures. It is noticed that the WOADL-BDDC model has
gained an accuracy of 94.38%, percent of 96.43%, sense of
85.28%, species of 95.93%, of 89.35%, and FPR of
4.07%.

TABLE I. RESULT ANALYSIS OF THE WOADL-BDDC SYSTEM WITH

DISTINCT MEASURES

Metrics TR set TS set

Accuracy 94.38 94.87
Precision 96.43 97.62
Sensitivity 85.28 88.89
Specificity 95.93 94.87
Fig. 3. (a) Original, (b) Pre-processed, (c) Segmented.
F1-Score 89.35 92.10
Fig. 3 visualization of the sample outcomes offered by the False Positive Rate 04.07 05.13
WOADL-BDDC system. Fig. 3(a) depicts the original brain
MRI and the pre-processed image is provided in Fig. 3(b).
Besides, the segmented images by the WOADL-BDDC
system are shown in Fig. 3(c).

Fig. 5. Result in the analysis of the WOADL-BDDC system under the TR

set.

Fig. 4. (a) TR set confusion matrix (b) TS set confusion matrix (c) TR set
PR-curve (d) TS set PR-curve.

302 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 1, 2023

A detailed ROC analysis of the WOADL-BDDC system

on the TS set is displayed in Fig. 8. The outcomes stated the
WOADL-BDDC approach had revealed its capability in
varying classifier classes.
The training accuracy ( ) and validation accuracy
( ) acquired by the WOADL-BDDC system on the test
database is depicted in Fig. 9. The simulation value stated that
the WOADL-BDDC method had reached increased values of
and . In particular, the looked that greater
than .
The training loss ( ) and validation loss ( )
accomplished by the WOADL-BDDC system under the test
database are displayed in Fig. 10. The simulation value
Fig. 6. Result in the analysis of the WOADL-BDDC system under the TS pointed out that the WOADL-BDDC system has achieved
set. lesser values of and . In particular, the is
lesser than .
Fig. 6 exemplifies a brief brain disorder classifier results of
the WOADL-BDDC approach under the TS set. The outcomes
referred that the WOADL-BDDC process has shown
enhanced results under all measures. It is observed that the
WOADL-BDDC system has gained an accuracy of 94.87%,
percent of 97.62%, sense of 88.89%, species of 94.87%,
of 92.10%, and FPR of 5.13%.
An obvious ROC analysis of the WOADL-BDDC
approach on the TR set is shown in Fig. 7. The outcomes
referred to the WOADL-BDDC algorithm have exhibited their
ability to categorize distinct classes.

Fig. 9. Result analysis of WOADL-BDDC system in terms of and

.

Fig. 7. ROC curve of WOADL-BDDC system under TR set.

Fig. 10. Result Analysis of the WOADL-BDDC model in terms of and

Fig. 8. ROC curve of WOADL-BDDC system under TS set. .

303 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 1, 2023

TABLE II. COMPARATIVE ANALYSIS OF THE WOADL-BDDC SYSTEM

WITH RECENT SYSTEMS

Methods Accuracy (%) Specificity (%)

WOADL-BDDC 94.38 95.93
Vgg16 90.35 89.50
Graph CNN 94.00 93.02
Modified ResNet18 92.81 94.66
Non-linear SVM 75.58 75.63
ResNet50-SVM 92.49 87.36
ResNet50-RF 86.31 79.57

A comprehensive comparison study of the WOADL-

BDDC model on brain disorder classification is shown in
Table II [23, 24].
Fig. 11 represents a comparative investigation of the
Fig. 12. analysis of WOADL-BDDC system with recent algorithms.
WOADL-BDDC technique with recent approaches in terms of
accuracy. The results indicated that the non-linear SVM and
ResNet50-RF models had attained the least accuracy values of These results confirmed the automated and accurate brain
75.58% and 86.31%, respectively. Following, the VGG16 has disorder classification results of the WOADL-BDDC model.
shown a slightly enhanced accuracy of 90.35%. In contrast, V. CONCLUSION
the graph CNN, modified ResNet18, and ResNet50-SVM
models have reached reasonable accuracy values of 94%, In this study, we have established a novel WOADL-BDDC
92.81%, and 92.49%. Finally, the WOADL-BDDC model has technique for recognizing brain disorders on MRI images.
obtained promising performance with a higher accuracy of Primarily, the presented WOADL-BDDC technique performs
94.38%. pre-processing in two stages: GF-based noise elimination and
U-Net segmentation-based skull stripping. Besides, the
Fig. 12 signifies a comparative investigation of the QuickNAT with RMSProp optimizer is applied to segment the
WOADL-BDDC system with current methods in terms of brain parts. For feature extraction, the WOADL-BDDC
species. The outcomes indicated that the non-linear SVM and technique performs a radiomics approach that separates
ResNet50-RF approaches had correspondingly accomplished features of all aspects. Finally, WOA with CRNN model is
the least values of 75.63% and 79.57%. Next, the applied for brain disorder classification. The experimental
VGG16 has demonstrated a somewhat improved of evaluation of the WOADL-BDDC technique is tested on
89.50%. In contrast, the graph CNN, modified ResNet18, and ADNI 3D dataset. The comprehensive comparison study
ResNet50-SVM models have reached reasonable values highlighted the adequate performance of the WOADL-BDDC
of 93.02, 94.66%, and 87.36%. Finally, the WOADL-BDDC technique in different measures. In the future, an ensemble of
methodology has attained promising performance with a DL-based classifiers can be employed to improve the
higher of 95.93%. WOADL-BDDC technique's detection performance.
VI. FUNDING STATEMENT
The authors received no specific funding for this study.
VII. CONFLICTS OF INTEREST
The authors declare they have no conflicts of interest to
report regarding the present study.
REFERENCES
[1] AlSaeed, D.; Omar, S.F. Brain MRI Analysis for Alzheimer’s Disease
Diagnosis Using CNN-Based Feature Extraction and Machine Learning.
Sensors , 22, 2911,2022. https://doi.org/10.3390/s22082911.
[2] Altaf, T., Anwar, S.M., Gul, N., Majeed, M.N. and Majid, M., Multi-
class Alzheimer's disease classification using image and clinical
features. Biomedical Signal Processing and Control, 43, pp.64-74,2018.
[3] Alwajih, R., Abdulkadir, S.J., Al Hussian, H., Aziz, N., Al-Tashi, Q.,
Mirjalili, S. and Alqushaibi, A., Hybrid binary whale with harris hawks
for feature selection. Neural Computing and Applications, pp.1-19,2022.
[4] An, N., Ding, H., Yang, J., Au, R. and Ang, T.F. Deep ensemble
learning for Alzheimer's disease classification. Journal of biomedical
informatics, 105, pp.103-112,2020.
Fig. 11. analysis of WOADL-BDDC system with recent algorithms.

304 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 1, 2023

[5] Basheera, S. and Ram, M.S.S.,Convolution neural network–based validation of an interpretable deep learning framework for Alzheimer’s
Alzheimer's disease classification using hybrid enhanced independent disease classification. Brain, 143(6), pp.1920-1933,2020.
component analysis based segmented gray matter of T2 weighted [18] Roy, A.G., Conjeti, S., Navab, N., Wachinger, C. and Alzheimer's
magnetic resonance imaging with clinical valuation. Alzheimer's & Disease Neuroimaging Initiative, QuickNAT: A fully convolutional
Dementia: Translational Research & Clinical Interventions, 5, pp.974- network for quick and accurate segmentation of
986,2019. neuroanatomy. NeuroImage, 186, pp.713-727,2019.
[6] Bi, X. and Wang, H., Early Alzheimer’s disease diagnosis based on EEG [19] Sampath, R. and Baskar, M., 3D brain image‐based Alzheimer's disease
spectral images using deep learning. Neural Networks, 114, pp.119- detection techniques using fish swarm optimizer's deep convolution
135,2019. Siamese neural network. Expert Systems, p.e12963,2022.
[7] Fouad, A., Moftah, H. M., &Hefny, H. A., Brain diagnoses detection [20] Shanmugam, J.V., Duraisamy, B., Simon, B.C. and Bhaskaran, P.,
using whale optimization algorithm based on ensemble learning Alzheimer’s disease classification using pre-trained deep
classifier. Int. J. Intell. Eng. Syst, 13(2), 40-51,2020. networks. Biomedical Signal Processing and Control, 71,
[8] Hu, Y., Wen, C., Cao, G., Wang, J. and Feng, Y., Brain network p.103217,2022.
connectivity feature extraction using deep learning for Alzheimer's [21] Sharma, R., Goel, T., Tanveer, M. and Murugan, R.,. FDN-ADNet:
disease classification. Neuroscience Letters, 782, p.136673.2022. Fuzzy LS-TWSVM based deep learning network for prognosis of the
[9] Jo, T., Nho, K., Bice, P., Saykin, A.J. and Alzheimer’s Disease Alzheimer’s disease using the sagittal plane of MRI scans. Applied Soft
Neuroimaging Initiative, Deep learning-based identification of genetic Computing, 115, p.108099,2022.
variants: application to Alzheimer’s disease classification. Briefings in [22] Tian, J., Smith, G., Guo, H., Liu, B., Pan, Z., Wang, Z., Xiong, S. and
Bioinformatics, 23(2), p.bbac022,2022. Fang, R. Modular machine learning for Alzheimer's disease
[10] Jo, T., Nho, K., Risacher, S.L. and Saykin, A.J., Deep learning detection classification from retinal vasculature. Scientific Reports, 11(1), pp.1-
of informative features in tau PET for Alzheimer’s disease 11,2021.
classification. BMC bioinformatics, 21(21), pp.1-13,2020. [23] Wang, Y., Liu, W., Yu, Y., Liu, J.J., Xue, H.D., Qi, Y.F., Lei, J., Yu,
[11] Arulananth, T.S., Balaji, L., Baskar, M. , PCA Based Dimensional Data J.C. and Jin, Z.Y.,CT radiomics nomogram for the preoperative
Reduction and Segmentation for DICOM Images. Neural Process Lett prediction of lymph node metastasis in gastric cancer. European
2020. https://doi.org/10.1007/s11063-020-10391-9. radiology, 30(2), pp.976-986,2020.
[12] Khagi, B., Kwon, G.R. and Lama, R., Comparative analysis of [24] Zhang, J., Zheng, B., Gao, A., Feng, X., Liang, D. and Long, X., A 3D
Alzheimer's disease classification by CDR level using CNN, feature densely connected convolution neural network with connection-wise
selection, and machine‐learning techniques. International Journal of attention mechanism for Alzheimer's disease classification. Magnetic
Imaging Systems and Technology, 29(3), pp.297-310.,2019. Resonance Imaging, 78, pp.119-126,2021.
[13] Khan, M.A., HCRNNIDS: hybrid convolutional recurrent neural [25] Amin, J., Sharif, M., Haldorai, A., Yasmin, M., and Nayak, R. S. , Brain
network-based network intrusion detection system. Processes, 9(5), tumor detection and classification using machine learning: a
p.834,2021. comprehensive survey. Complex Intell. Syst. 8, pp.3161–3183,2021.
[14] Li, L. and Ma, H., Pulse coupled neural network-based multimodal doi: 10.1007/s40747-021-00563-y.
medical image fusion via guided filtering and WSEML in NSCT [26] Baskar, M., Renuka Devi, R., Ramkumar, J. , Region Centric Minutiae
domain. Entropy, 23(5), p.591,2021. Propagation Measure Orient Forgery Detection with Finger Print
[15] Liang, G., Xing, X., Liu, L., Zhang, Y., Ying, Q., Lin, A.L. and Jacobs, Analysis in Health Care Systems. Neural Process Lett . 2021.
N., Alzheimer’s disease classification using 2d convolutional neural https://doi.org/10.1007/s11063-020-10407-4.
networks. In 2021 43rd Annual International Conference of the IEEE [27] Houssein, E. H., Ibrahim, I. E., Neggaz, N., Hassaballah, M., and
Engineering in Medicine & Biology Society (EMBC) (pp. 3008-3012). Wazery, Y. M. , An efficient ecg arrhythmia classification method based
IEEE,2021. on manta ray foraging optimization. Expert. Syst. Appl. 181,
[16] Odusami, M.; Maskeliunas, R.; Damasevicius, R.; Krilavicius, T. 115131,2021. doi: 10.1016/j.eswa.2021.115131.
Analysis of Features of Alzheimer’s Disease: Detection of Early Stage [28] Bekhet, S., Hassaballah, M., Kenk, M. A., and Hameed, M. A. , An
from Functional Brain Changes in Magnetic Resonance Images Using a artificial intelligence based technique for COVID-19 diagnosis from
Finetuned ResNet18 Network. Diagnostics 2021, 11, 1071. https:// chest x-ray, in 2020 2nd Novel Intelligent and Leading Emerging
doi.org/10.3390/diagnostics11061071. Sciences Conference (NILES) (Giza: IEEE), pp.191–195,2020.
[17] Qiu, S., Joshi, P.S., Miller, M.I., Xue, C., Zhou, X., Karjadi, C., Chang,
G.H., Joshi, A.S., Dwyer, B., Zhu, S. and Kaku, M., Development and

305 | P a g e
www.ijacsa.thesai.org