Volume 8, Issue 1, January – 2023 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

An Intelligent System to Analyze the Functional

Magnetic Resonance Imaging fMRI
George Karraz 12
1
Department of Artificial Intelligence & Natural Language Processing, Faculty of Information Technology Engineering,
Damascus University.
2
Department of Informatics, Faculty of Engineering, Al-Sham Private University, Damascus, Syria.

Abstract:- A Lot of medical projects aim to combine I. INTRODUCTION

biology with computer science like artificial limb which
is able to simulate real limb's activities to some extent, Japanese S.Ogawa [16] was the first to discover the
and that requires to comprehend the neurological map of technique of functional magnetic resonance imaging (fMRI)
the brain. The best way to measure the brain's activity is in 1990. Ogawa discovered that the magnetic properties of
Functional Magnetic Resonance Imaging (fMRI), where blood vary with the percentage of oxygen in it. At the end of
it is a functional neuroimaging procedure using MRI his research, Ogawa predicted that this method could be
technology that measures brain activity by detecting used to measure brain activity by taking advantage of the
changes associated with blood flow. In this paper we difference in the amount of oxygen in the blood when cells
develop an automatic system based on soft computing perform a certain activity compared to other cells. Ogawa
methods, to analyze fMRI Images and conclude their has done more research to prove that his method can be
proper intended behavior. Our data was composed from successful in measuring brain activity, shared by many
two parts, the major part was obtained from the famous scientists in the world. Indeed, the effectiveness of this
dataset (A test-retest fMRI dataset for motor, language method has been proven and developed to become the most
and spatial attention functions), which has a important tool in revealing the secrets of the brain.
representation of five different behaviors “finger foot
and lip movement, overt verb generation, covert verb In the last ten years, the use of this technology has
generation, overt word repetition and landmark tasks”, increased significantly, especially in the field of studying the
where the second part was prepared by us using images brain and its functions and related research. Studying the
that free downloaded from internet network. Our functions of the brain and how it functions is a difficult
developed automatic classification system is based on challenge, because the brain is fortified within strong bones
neural network framework, which is proceeding in two for its protection (the bones of the skull). Scientists have
stages: developed many ways to study the brain, but fMRI
technology may be the most important of these methods
1. The first stage extracts four specific features, because it is safe and does not include any radiation or any
through applying sophisticated techniques for automatic interference surgical. In addition to the research applications
image processing and analysis, related to the presence of of fMRI, its use in the clinical field in hospitals and its
different intensity values and their addresses over the 2 applications in them is increasing, especially in pre-surgery
dimensions studied images. The selected features were cases of the brain.
unique and contribute to make our system, good
We know a lot about the brain before the advent of
represented.
fMRI using different methods of studying the brain, perhaps
2. The second stage is a classification technique, the oldest is postmortem autopsy. For example, Broca's area
through designing a suitable artificial intelligence system of the brain (responsible for speech and pronunciation) was
architecture and learning algorithm. We did a lot of discovered by autopsying the brain of a person who was
experiments in order to select the best neural network healthy, but then contracted a disease (what is known today
architecture and training method, the experiments as a stroke) and lost the ability to speak. The brain of this
proved that the best performance was achieved in three person was dissected after his death by the doctor Paul
layers neural network: input, hidden and output layers, Broca [1] and it was discovered that there was a death of
with a training method based on Back propagation cells in the brain in a certain area, the doctor concluded that
algorithm, and sigmoid activation function. Developed this area is responsible for speech, after that indeed
system achieved an accuracy of 94.4%. subsequent experiments proved the correctness of his words
and this area was named after Broca’s area. The brain is
Keywords:- fMRI; Neural Networks; Brain Activity made up of several regions that have different and varied
Automatic Interpretation; Fuzzy C-maen clustering; Linear functions. Research on brain functions is still trying to
Regression. understand the brain, and although we know a lot about the
brain and its secrets, there is also a lot that we don't know.
Let's take another famous example,Wernicke's area [2],
which is a region of the brain responsible for understanding
speech. For example, any disease in Wernicke's area may
result in the patient speaking in an incomprehensible

IJISRT23JAN1316 www.ijisrt.com 1672

Volume 8, Issue 1, January – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
language. He may speak well because Broca's area is not fMRI results. For example, Broca's area alone cannot
affected, but his words have no meaning. produce speech, but we need other areas of the brain
responsible for the movement and coordination of muscles
The brain is interconnected with each other and there in the face, wemay also need the area responsible for
is no region that can perform its functions on its own, so a memory toretrieve memorized words from it. Figure 1
specific region of the brain may have multiple functions, or illustrates the functional division of the surface cerebral
the function may be performed by several regions of the cortex as it reported by H. Whitaker [10].
brain. We must therefore be careful when interpreting the

Fig. 1: The functional division of the surface cerebral cortex

II. RELATED WOEKS machine analysis with a least absolute shrinkage and
selection operator to evaluate five performance metrics:
C. Candemir [5] presented a two-fold innovative study. accuracy, recall, and specificity. Principal component
First, detection algorithms to locate the activations in fMRI features showed the best classification performance in all
signals. Furthermore, proposed and compared a set of aspects of metrics compared to BOLD response to single-
change points (CP) analysis methods, such as a regression- task fMRI. The approach showed better performance in
based method (RBM), a statistical method (SM), and a discriminating patients with PI from HCs, compared to
method that based on the mean difference of double sliding single-task fMRI.
windows (MDSW)) to locate such points (CP).
H. Hasan [9], analyzed the underlying causes of the
Secondly, these methods were applied on the fMRI limited performance of the projected density algorithm when
signals, which are acquired from the real subjects, while applied to brain data. In addition, compared it with an
they were performing fMRI tasks. Proposed methods were approach that relies on the optimization of the conductivities
applied to three different fMRI experiments with a motor, a of a small number of tissue compartments of anatomically
visual and a linguistic tasks. The analysis shows that the detailed head models, which reconstructed from structural
methods find activations in accordance with established MR data. Both for simulated ground truth data and magnetic
techniques such as statistical parametric maps (SPM). The resonance current density imaging MRCDI data, the
acquired results showed accuracy up to 94 %, also showed obtained results indicated that the estimation of densities
that the proposed techniques can be used effectively to benefits more than using a personalized volume conductor
locate the activation times on fMRI time series. model. In particular, they introduced a hierarchical statistical
testing approach as a principled way to test and compare the
M. H. Lee [15], investigated the differential spatial quality of reconstructed current density images that accounts
covariance pattern of blood oxygen level-dependent for the limited signal-to-noise ratio of the human in-vivo
(BOLD) responses to single-task and multitask functional MRCDI data and that the ground truth of the current density
magnetic resonance imaging (fMRI) between patients with is unknown for measured data. The results indicated that the
psychophysiological insomnia (PI) and healthy controls statistical testing approach constitutes a valuable framework
(HCs), and evaluated features generated by principal for the further development of accurate volume conductor
component analysis (PCA) for discrimination of PI from models of the head. The findings also highlight the
HC, compared to features generated from BOLD responses importance of tailoring the reconstruction approaches to the
to single-task fMRI using machine learning methods. In 19 quality and specific properties of the available data.
patients with PI and 21 HCs, the mean beta value for each
region of interest (ROIbval) was calculated with three K. J. Gorgolewski [12], found that the adaptive
contrast images (i.e., sleep-related picture, sleep-related threshold can improve reliability, mainly by accounting for
sound, and stroop stimuli). They performed discrimination the general signal variance. This in turn increases the
analysis and compared with features generated from BOLD likelihood that the true activation pattern can be determined
responses to single-task fMRI. They applied support vector

IJISRT23JAN1316 www.ijisrt.com 1673

Volume 8, Issue 1, January – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
offering an automatic yet flexible way to threshold single the activated zones in the image, we ensure that the main
subject fMRI maps. reason for increasing accuracy of image interpretation
isrelated mainly, to the two developed techniques of image
B.A. Kirchhoff [4] had adopted a study that reveals automatic analysis.
functional anatomic correlates of verbal and perceptual
strategies that are variably used by individuals during V. PROPOSED WORK
encoding. These strategies engage distinct brain regions and
may separately influence memory performance. A. fMRI Images Processing
The fMRI image was represented in 2D and in grayscale
K. J. Duncan [11] proposed a technique based on a mode using the functions designated for this purpose in
statistical methodology, aiming to evaluate the consistency, Matlablanguage, then we developed two different adaptive
associated with functionally localizing reading- and object- approaches to analyze automatically the studied fMRI
sensitive in areas of left occipital-temporal cortex. The images in our dataset, in order to isolate the activated zones
obtained results were closely match previous studies with from other details presented in the images.
peak activations located in the posterior occipital-temporal
sulcus according to the written words. Then concluded that The 1st approach is an intelligent model based on
intra-subject variability was surprisingly high, with between linear regression that operates to determine an adaptive
one third and three quarters of the voxels in a given image appropriate threshold according to the gray intensity, in an
not corresponding to those activated in the main task. This attempt to facilitate the isolation of active zones using this
level of variability stands in striking contrast to the type of medical imaging. Through our observations to the
consistency seen in retinotopically defined areas and has images presented in the studied dataset, we concluded that
important implications for designing robust but efficient it’s very important to use the statistics features that could be
functional localizer scans. calculated over all of image intensity values (maximum,
mean and standard deviation), this technique gave us the
G. Ganis [8] found that fMRI revealed that well- possibility to calculate an important statistical factor (X) that
rehearsed lies that fit into a coherent story elicit more affects in estimating the threshold in the developed
activation in right anterior frontal cortices than spontaneous approach, so the linear regression model takes X as input,
lies that do not fit into a story, whereas the opposite pattern where the output is the threshold value.
occurs in the anterior cingulate and in posterior visual
cortex. The 2nd approach based on the using of Fuzzy C-
means Clustering algorithm (fcm), in order to determine the
D. K.Jones [7] had shown through their study that it is centroids for every cluster of intensity values in fMRI
possible to obtain robust and high quality diffusion tensor image, It's clearly noted that every activated zone
MR data at 1.5 Tesla with isotropic resolution (2.5 × 2.5 × correspond to a specific centroid and cluster, here we used
2.5 mm) from the whole brain within a sufficiently short the Z-Scores as a statistical representation, to specify the
imaging time that it may be incorporated into clinical number of fMRI image clusters which are needed in fcm
imaging protocols. operation. In our two developed approaches we extract
coordinates of detected activated zones in fMRI images, in
III. PROBLEM DICRIPTION order to achieve our target, in classifying the presented
behaviors in the brain regions.
There have been many serious attempts to analyze
(fMRI) images, to isolate the active zones presented in the  Linear Regression Model:
image, in order to proceed an automatic interpretation of The linear regression model [6] is our first approach to
what is associated with these zones of behaviors according reveal the correct adaptive threshold to isolate the activation
to their locations within the map of the brain image. The zones in fMRI image, firstly we divided our dataset in two
problem of our research is to use more of effective parts, the first part forms 60% of total images and the second
techniques in the fields of analysis and automatic one forms 40%, the first part forms training dataset, and the
interpretation of the studied image, based on the artificial second part forms the testing dataset, the process began to
intelligence algorithms, to achieve, as possible as, better represent all 2D images in our database graphically in 3D, in
results in accuracy of our developed system, in comparison order to illustrate the intensity values of studied image on
to our peer researches. Z-axis, and the coordinates of image on X-axis and Y-axis,
Figure 2 illustrates the fMRI image and the manner of its
IV. MOTIVATION AND OBJECTIVE representation.
The main motivation behind this research is to present
Then we created a target vector corresponds to
a new initiative in the subject of automated analysis of fMRI threshold values of the activated zones in images, for every
images, in the first stage of research, using two parallel image we specified an adaptive threshold, the created vector
mechanisms of artificial intelligence algorithms, where the (Y) represents the regression model target in the training
developed mechanisms showed equal high efficiency in stage, the input of model (X) is a vector of features, that can
separating activation zones in the studied images. In be calculated from three specified statistical values extracted
addition, the two developed mechanisms have greatly
from all studied fMRI images, (maximum (MAX), average
contributed to support the second stage of our research,
(MEAN) and standard deviation (STD)), as illustrated in
which is the automatic interpretation of behaviors related to
Equation 1

IJISRT23JAN1316 www.ijisrt.com 1674

Volume 8, Issue 1, January – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
MAXVal − Meanval The third element in linear regression model is the
X= (1) optimization method, we use the Gradient Descent (GD)
STDVal
algorithm in order to minimize the cost function and find the
best hypotheses that reveals the accurate solution of our
problem to determine the adaptive threshold in order to
isolate the activation zones in fMRI image. The (GD)
function is illustrated in Equation 4.

∂
θj = θj − α J (θ0 , θ1 )
∂θj

for j = 0 and 1 (4)

Where: α is the learning rate which may be selected to
be adaptive to the input and output values, the best value of
α in our system was 0.01, at the beginning of training
process the system suggest initial values to θ_0 and θ_1 to
form an initial hypothesis, then the system enters in
Fig. 2: fMRI image and its 3D representation iterations epochs to minimize the error or cost function
value calculated in every iteration, the system will stopped
Our linear regression model (LR) depend on three main when it reaches a specified value of error or a determinate
elements, the first element is the hypothesis of linear number of iterations. The Figure 3 below illustrates the
function which is illustrated in Equation 2. performance of optimization method, the best performance
is MSE=0.0036 at the 25th iteration.
h(x) = θ0 + θ1 x (2)
This stage of research was concluded with a linear
Where h(x), x,θ0 and θ1 represent the estimated output,
function that used to produce an adaptive threshold to isolate
input, weights of linear model, respectively.
the active zone in any studied fMRI image. We test the
The second element in linear regression model is the developed model on our testing dataset and realized an
cost function which is used in optimization process, in this accuracy is about 96.28 %, the accuracy is calculated as a
model we used the mean squared error (MSE) as a cost ratio between the numbers of selected activated zones (SAZ)
function as reported in Equation 3. and real activated zones (RAZ) see the following Equation
5.
𝑚
1
𝐽(𝜃0 , 𝜃1 ) = ∑(ℎ𝜃 (𝑥 (𝑖) − 𝑦 𝑖 )2 (3)
2𝑚
𝑖=1

Fig. 3: Performance of optimization method (minimized the MSE by GD algorithm)

SAZ
ACC = (5)
RAZ

Figure 4 below in its part (2) illustrates the result of applying the produced adaptive threshold on the studied image in part (1)

IJISRT23JAN1316 www.ijisrt.com 1675

Volume 8, Issue 1, January – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

Fig. 4: Result of applying a produced adaptive threshold (2) on a studied image (1), by using LR algorithm

 Fuzzy C-mean Clustering the algorithm updates its centroids by measuring the mean
Fuzzy C-means Clustering algorithm (FCM) [14] is an of every produced cluster, and iterates to minimize the error,
our second approach to reveal the correct adaptive threshold the process will be terminated in case of reached the desired
to isolate the active zones in fMRI image, the first step is to error or the a selected number of iterations. Let we have n
represent the studied fMRI image using statistical Z-score pixels in the image and we want to classify the pixels in k
basing on two statistical values of image, which are the clusters, so the distance (Dk) between n pixels (P) and
mean value (MEAN) and the standard deviation (STD) centroid (Ck) is calculated as illustrated in Equation 7 below:
using the Equation 6.
Dk = ∑(Pi − Ck )2 (7)
Image(i,j)− MEAN(image)
Z(i,j) = (6) i
STD(image)
Where: i varies from 1 to n
Where: i, j varies from 1 to the number of rows and
columns in image map, respectively. Error is calculated as illustrated in Equation 8
k
The statistical Z-score gave us the possibility to
estimate the number of clusters K that the intensity values of Err = ∑ Dk (8)
image belong to it, we considered K as an input to FCM j=1
algorithm, which operates a set of iterations to classify all
the pixels in image according to clusters centers (centroids), We obtained an important result from applying this
the algorithm begins by determining initial centroids and it approach is to determine the cluster of activation zones in
calculates the distances between all centroids and image fMRI image and their addresses in the image map. The
pixels, every pixel will belong to a cluster that realizes the outcome accuracy of this approach was calculated according
minimum distance between its centroid and the pixel, the to the Equation 5 that previously mentioned, and is reached
error in every iterationaction represents the sum of squared 96.7. Figure 5below illustrates the performance of fcm on a
distances between cluster centroids and the pixels of image, studied fMRI image.

Fig. 5: Performance of fcm onfMRI image

IJISRT23JAN1316 www.ijisrt.com 1676

Volume 8, Issue 1, January – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
B. Feature Extraction
After isolating the active regions infMRI images, we C. Target Preparation
have extracted four characteristic features over all images in Our target represent the output of our developed
our training dataset: automatic classification system in the training stage, it is 2D
 The projection pattern (PP) of the studied fMRI image, matrix (5 x 5040000) of elements, according to the number
which is a Vector represents the image in two projections, of five activities adopted in our dataset which are finger foot
the Vertical and horizontal, the values in this feature takes and lip movement (FFLM), overt verb generation (OVG),
one of two values, 1 for vertical and 0 for horizontal covert verb generation (CVG), overt word repetition (OWR)
projection. and landmark tasks (LT), and according to the number of
 Activation Score (AS) which it was represented in a pixels in all images of training dataset. Every vector in the
vector of values that take one of two values 1 in case of target represent an activity task from the five studied
activation and 0 in case of non-activation. activities, every element in every vector takes one of two
 The addresses of pixels in fMRI image on X-axis of Image values: 1 or 0 in cases of activation and non-activation
map (Xind). respectively.
 The addresses of pixels in fMRI image on Y-axis of Image D. Automatic Classification System
map (Yind).
 Architecture
These features were arranged in a 2D matrix of size (4 The automatic classification system was build based on a
x 5040000) according to the amounts of pixels and features neural network contains of three layers: the input layer has
in our training dataset that have 65 fMRI images. Features four nodes corresponding to the number of studied features,
matrix will be the input of our developed automatic hidden layer has six nodes and output layer has five nodes
classification system in the training stage. All the features corresponding to the classified behaviors. Figure 6 below
vectors was normalized in everystudied image to take values illustrates the system architecture.
in the range [0 1], both in training and testing stages.

Fig. 6: Developed neural network architecture

 Learning Method It is also worth noting that when data passes through the
We tried many frameworks to learn our neural network three layers, it is processed to result in what is called the
system, Bayesian, Backpropagation (Bp) [3], Quasi Newton error function, which is the result of the difference between
and Levenberg Marquardt, our experiments and evaluations the expected (desired) outputs and the actual (real) outputs,
proved that the performance of BP reached the best and this output is called the error ratio and reaching it is the
performance on our testing dataset. desired goal of using the algorithm. The Bp algorithm can
summarized as illustrated in Diagram 1.

IJISRT23JAN1316 www.ijisrt.com 1677

Volume 8, Issue 1, January – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Firstly the algorithm calculates the deltas from hidden
1. Initialize weights (typically random) layer to output layer, see Equations 9, 10 and 11 below.
2. Keep doing epochs delta wji = α Ai (Ti − Oi ) Der(Oi ) (9)
a. For each example e in training set do:
i. Forward pass to compute 1
 O = neural- network –output Due to Oi =
1 + e−Ai
 T: is the real output
 .miss = (T-O) at each output unit Der(Oi ) = Oi (1 − Oi ) (10)
ii. Backward pass to calculate deltas to
weights delta Wji = α Ai (Ti − Oi )Oi (1 − Oi ) (11)
iii. Update all weights
b. End. Then the algorithm calculates deltas from input layer
3. Until tuning set error stops improving. to hidden layer, see the Equations 12 and 13.

Diag.1: Bp algorithm steps missj = ∑[Ai (1 − A𝑖 )(T𝑖 − Wji ) (12)

𝑖
It begins to compute deltas to weights from hidden
layer to output layer, we used the derivation of activation
delta ki = α Ik Aj (1 − Aj ) missj (13)
function which is the logarithm sigmoid, the real output (T),
the neural network output (O), the output of input layer (A),
the input of input layer (I) and the learning rate (α) which The learning algorithm achieved the best solution
chosen in our system equals to 0.01, let indicate the symbols through 550 iteration with error about 0.0048, Figure 7
(i), (j), (k) to the nodes of output, hidden and input layers, below illustrates the performance of backpropagation
in our system i, j and k varies from 1 to 5, 1 to 6 and 1 to 4 algorithm during the learning stage.
respectively.

Fig. 7: Performance backpropagation learning algorithm

 Testing and Results We fixed an adaptive threshold to classify the values in

The developed system was tested by using testing the output vector, we consider the values >=0.5 represent
dataset which represents 40% of our database, the testing positive cases and < 0.5 represent negative cases.TABLE I
dataset was not involved in learning stage, for every image, illustrates the results of testing stage. We used the symbols
we extract four vectors of features according to the input Cp , Cn , Tp , Tn , Fp and and Fn to represent the truly real
layer of neural network. positive, truly real negative, truly positive detected, truly
negative detected, false positive detected and false negative
The performance of system in testing stage was detected cases respectively.
excellent, it realized a sophisticated capability to recognize
43 cases of behaviors from total 45 cases and it gave only 3
false alarms (false positive cases), in other hand it gave only
2 false negative cases.

IJISRT23JAN1316 www.ijisrt.com 1678

Volume 8, Issue 1, January – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Cp Cn Tp Tn FP Fn
45 45 43 42 3 2
Table 1: THE RESULTS OF TESTING STAGE

According to the results reported in TABLE I, we used Tp

a statistical functions to calculate the sensitivity (Se ), Se = (14)
Cp
specificity (Sp ) and accuracy (Acc ) these functions represent
the system capacity to recognize correctly the positive cases, Tn
negative cases and both of them respectively, where the Sp = (15)
Cn
error( Err ) represents the probability of both false positive
and negative cases, see Equations 14, 15, 16 and 17 TP + Tn
illustrate how we calculate Se,Sp ,Acc and Err .The TABLEII Acc = (16)
Cp + Cn
reported the results of statistical evaluation process.
Err = 1 − Acc (17)

Se Sp Acc Err
0.956 0.933 0.944 0.056
Table 2: The RESULTS OF STATISTICAL EVALUATION PROCESS

One of the efficient statistical methods to evaluate our classifier, the obtained results proved that the best threshold
results is using the receiver operating characteristics curve is 0.5, <0.5 for the negative cases and >=0.5for the positive
(ROC) which represents the relationship between the cases, see TABLE III, then we plotted ROC curve as
probabilities of true detected positive cases (Se ) and the illustrated in Figure 8, we also calculated the area under
probabilities of false positive detected cases (1- Sp ), the ROC (AUC) as an important index to evaluate the ROC
probabilities used in ROC curve is calculated according to curve efficiency, the AUC represents statistically the mean
five thresholds, in the range [0 1]. we calculated according of obtained sensitivities at the selected thresholds, AUC in
to every selected threshold the sensitivity and specificity of our system is 0.969.

Threshold Se Sp 1-Sp
0.1 1 0.60 0.40
0.2 0.978 0.64 0.36
0.3 0.978 0.71 0.29
0.5 0.956 0.933 0.067
0.6 0.933 0.978 0.022
Table 3: SENSITIVITIES AND SPECIFICITIES OF ROC

Fig. 8: ROC Curve

Figure 9 below illustrates all the steps of our developed system on a studied fMRI image, the image has an activated zone
represents FFLM behavior, it is noted that the system classified clearly this behavior according to the activated zone in the image.

IJISRT23JAN1316 www.ijisrt.com 1679

Volume 8, Issue 1, January – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

Fig. 9: Application of developed system on a studied fMRI image with an activated zone, the red line represents the threshold
value of classification, FFLM behavior was recognized clearly

VI. DICUSSION VII. DATA AVAILABILITY

Through our methodology of research, we were able to Our major and first part of our dataset was obtained
introduce new techniques related to the field of processing from the famous database (A test-retest fMRI dataset for
and automatic analysis of digital images in order to extract motor, language and spatial attention functions), which
the accurate information required to be known from the contains 100 fMRI images with accordance to five different
studied images, which were represented by fMRI images, behaviors, this database is available free in [13], in the
this process facilitated to identify successfully and second part, we added other free 12 fMRI images obtained
accurately the behaviors represented in the studied images, from Google, that were documented and classified
which is the desired goal of this research. manually in our dataset, for determining the behaviors
presented in the added images and correlated to behaviors
The results were obtained demonstrated a significant presented in our first part of dataset.
improvement in performance, compared to the results
documented in the scientific research literature relevant to VIII. FUNDING STATEMENT
our research topic. We hope to enrich our research balance
with more data in addition to the data that was used in the There is no funding for the research done, it is an
research, so our developed techniques become more individual effort by the author.
generalized.

IJISRT23JAN1316 www.ijisrt.com 1680

Volume 8, Issue 1, January – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
REFERENCES [16.] L.Szilágyi, L.Lefkovits and D.Iclanzan, “A review on
suppressed fuzzy c-means clustering models”, Acta
[1.] Gaur, A. Augustyn, A. Zeidan, et,al. “Broca area”, Universitatis Sapientiae, Informatica, Hungarian
Encyclopaedia Britannica, Inc, Dec 15, 2022. University of Transylvania,Hungary, vol.12, no.2,
[2.] https://www.britannica.com/science/Broca-area pp.302-324, Sep 27, 2021.
[3.] Gaur, A. Augustyn, A. Zeidan,et,al. “Wernicke area”, [17.] M. H. Lee, Nambeom Kim, Jaeeun Yoo, et.al,
Encyclopaedia Britannica, Inc, March 10, 2020. “Multitask fMRI and machine learning approach
https://www.britannica.com/science/Wernicke-area improve prediction of differential brain activity
[4.] Singh, S. Kushwaha, M. Alarfaj and M. Singh, pattern in patients with insomnia disorder”, Scientific
“Comprehensive overview of backpropagation Reports, Nature Portfolio,Springer Nature, vol.11,
algorithm for digital image denoising”, Electronics, no.9402, 13 pages, Apr 30, 2021.
MDPI, vol.10, no. (11)1590, 15 pages, May 17, 2022. [18.] S. Ogawa, T. M. Lee, A. R. Kay and D. W. Tank,
[5.] A. Kirchhoff and L. Randy Buckner, “Functional “Brain magnetic resonance imaging with contrast
anatomic correlates of individual differences in dependent on blood oxygenation”,Proceedings of the
memory”, Neuron,Cambridge, USA, vol.51, no.2, National Academy of Sciences,USA, vol. 87, no 24,
pp.263-274, July 20, 2006. pp. 9868-9872, December 199
[6.] Candemir and K. Oğuz, “Change point detection
,methods for locating activations in functional
neuronal images”, BSEU Journal of Science, vol. 9,
no.1, pp 541-554, May 30. 2022.
[7.] D.H. Maulud and A. M. Abdulazeez, “A review on
linear regression comprehensive in machine
learning”, Journal of Applied Science and
Technology Trends, Kurdistan, vol. 01, no. 04, pp.
140 –147, 2020
[8.] K. Jones, SC. Williams, D. Gasston, et al., “Isotropic
resolution diffusion tensor imaging with whole brain
acquisition in a clinically acceptable time”, Hum
Brain Mapping, Wiley Online Library,vol.15, pp.216-
230, 2002.
[9.] G .Ganis, S. M. Kosslyn, S. Stose, W. L. Thompson
and D. A. Y. Todd, “Neural correlates of different
types of deception: An fMRI investigation”. Cerebral
Cortex, Oxford Academic, vol. 13, pp. 830-836,
August 2003.
[10.] H. Hasan, E. ğlu , O. Puonti , C. Göksu Thielscher, et
al., “On the reconstruction of magnetic resonance
current density images of the human brain: Pitfalls
and perspectives”, Neuroimage, Science Direct,
Elsevier,vol. 243, Article ID 118517, 15 pages, 2021.
[11.] H. Whitaker, “Concise Encyclopedia of Brain and
Language”, 1st Edition, e-Book ISBN:
9780080964997, Paperback ISBN: 9780081014516,
January 14, 2010.
[12.] K. J. Duncan, C. Pattamadilok, I .Knierim and JT
.Devlin, “Consistency and variability in functional
localizers”,Neuroimage, Science Direct, Elsevier, vol
46, pp. 1018-1026, July 15, 2009.
[13.] K. J. Gorgolewski, A. J. Storkey, M. E. Bastin and C.
R. Pernet, “Adaptive thresholding for reliable
topological inference in single subject fMRI
analysis”, Frontiers in Human
Neuroscience,Switzerland, vol. 6, Article ID 245, pp.
1-14, August 25, 2012.
[14.] K. J. Gorgolewski, A. Storkey1 , M. E Bastin , I. R.
Whittle, et al., “A test-retest fMRI dataset for motor,
language and spatial attention functions”,
Gigascience,Oxford Academic,London, UK,vol.2,
no.6, 4 pages, April 29 2013.
[15.] http://openfmri.s3.amazonaws.com/tarballs/ds000114
_R2.0.1_raw.zip

IJISRT23JAN1316 www.ijisrt.com 1681