=Paper=
{{Paper
|id=Vol-3105/paper33
|storemode=property
|title=Detection of Covid-19 From an Imbalanced Chest X-ray Image Data Set
|pdfUrl=https://ceur-ws.org/Vol-3105/paper33.pdf
|volume=Vol-3105
|authors=Sharath Madhu Manjunath,Manju Gurjar,Neil O'Kane,Andrew McCarren,Leonardo Gualano
|dblpUrl=https://dblp.org/rec/conf/aics/ManjunathGOMG21
}}
==Detection of Covid-19 From an Imbalanced Chest X-ray Image Data Set==
https://ceur-ws.org/Vol-3105/paper33.pdf
Detection of Covid-19 From an Imbalanced
Chest X-ray Image Data Set
Sharath Madhu Manjunath1[0000−0002−1894−5331] , Manju
2[0000−0002−6774−1087]
Gurjar , Neil O’Kane3[0000−0003−3506−1324] , Andrew
4[0000−0002−7297−0984]
McCarren , and Leonardo Gualano5[0000−0002−7543−4394]
1
Dublin City University, Dublin, Ireland
2
sharath.madhumanjunath2@mail.dcu.ie
Abstract. The Covid-19 pandemic has spread quickly, making identifi-
cation of the virus critically important in assisting overburdened health-
care systems. Numerous techniques have been used to identify Covid-19,
of which the Polymerase chain reaction (PCR) test is the most common.
However, obtaining results from the PCR test can take up to two days.
An alternative is to use X-ray images of the subject’s chest area as in-
puts to a deep learning neural networks algorithm. The two problems
with this approach are the choice of architecture and the method used
to deal with the imbalanced data. In this study a comparative analysis of
a standard convolutional neural network (CNN) and a number of trans-
fer learning algorithms with a range of imbalanced data techniques was
conducted to detect Covid-19 from a data set of chest x-ray images. This
data set was an amalgamation of two data sets extracted from the Kag-
gle Covid-19 open source data repository and non-Covid illnesses taken
from the National Institute of Health. The resulting data set was had
over 115k records and 15 different type of findings ranging from no-illness
to illnesses such as Covid-19, emphysema and lung cancer. This study
addresses the problem of class imbalance on the largest data set used
for x-ray detection of Covid-19 by combining undersampling and over-
sampling methods. The results showed that a CNN model in conjunction
with these random over and under sampling methods outperformed all
other candidates when identifying Covid-19 with a F1-score of 93%, a
precision of 90% and a recall of 91%.
Keywords: Covid-19 · Oversampling · Undersampling · CNN · transfer
learning · chest x-ray
1 Introduction
The novel coronavirus illness of 2019 has created a substantial and immediate
threat to global health. The outbreak has paralyzed the world economy and
created widespread disruptions to people’s lives. According to the World Health
Organization (WHO), the coronavirus pandemic is putting unprecedented and
escalating strain on healthcare systems around the world.
2 S. Manjunath et M. Gurjar
Covid-19 has been diagnosed using a range of medical imaging modalities,
blood tests (CBCs) and PCR. According to the WHO, all diagnoses of Covid-19
must be validated by reverse-transcription polymerase chain reaction tests. PCR
techniques are performed manually and have a high degree of inaccuracy with a
positivity rate of only 63% [1]. Also, PCR techniques require a lot of time and
resources. Patients must apply to health centers in batches due to the virus’s
fast-spreading nature [2]. These factors ensure that the necessity for quick diag-
nostic procedures is of critical importance at this point in time. Another tech-
nique for monitoring and diagnosing SARS-CoV-2 infections is visually, using
radiological images such as chest x-rays or computed tomography (CT). Covid-
19 creates abnormalities, evident on chest x-rays and CT scans, in the form of
ground-glass opacities. It is foreseeable that a diagnostic could be made with
radiological imaging. This would have certain advantages over PCR testing in
terms of the detection rate in the early phases of Covid-19. One way would be
to have professionals manually analyse the images. However, an alternative is an
Artificial Intelligence based approach, which could provide a more timely and
accurate means of diagnosing Covid-19.
Deep learning may be used to make predictions based on medical images by
extracting distinctive characteristics such as shape and spatial rotation from the
images [3]. CNNs have been critical in extracting features and learning patterns
that enable prediction [4].
For this study, a comparative analysis of a standard convolution neural net-
work (CNN) and number of transfer learning algorithms with a range of imbal-
anced data techniques was conducted to detect Covid-19 from a data set of chest
x-ray images. Each algorithm and imbalance combination will be assessed using
the F1-score, Precision, Recall and the area under the curve (AUC).
In section 2, past research is examined that employed deep neural networks
to identify Covid-19 and other chest illnesses. The data sets used are described
in section 3 and the methodology is detailed in section 4. Finally, our findings
and analysis based on the model and data set are presented in section 5.
2 Literature Review
Researchers have developed deep learning techniques to detect and diagnose
Covid-19 based on x-ray images of the chest. This literature review, describes
the development of deep learning techniques, to detect Covid-19 and the methods
used to handle an imbalanced data set.
2.1 Deep Neural Networks for detecting Covid-19 cases using chest
x-ray images:
A Generative Adversarial Network, is introduced by Loey et al., for pre-processing
chest x-ray images [5]. The scheme used three pre-trained models: AlexNet,
GoogleNet, and RestNet18. The data set included 69 Covid-19 images, 79 pneu-
monia bacteria images, 79 pneumonia virus images, and 79 normal images.
Detection of Covid-19 From an Imbalanced Chest X-ray Image Data Set 3
GoogleNet was selected as a main deep learning technique with a 80.6% test
accuracy containing four class scenarios, AlexNet with a 85.2% test accuracy
with three class scenarios and again GoogleNet with 99.9% scenario with two
class scenarios. This shows how results can be seen to improve when the number
of classes is reduced.
A combination of a CNN and a Long Short-Term Memory (LSTM), was
used by Islam et al., to detect Covid-19 using x-ray images [6]. In this system, a
CNN is used for feature extraction and LSTM for detection using the extracted
feature. The data set includes 613 Covid-19 images, 1525 pneumonia images, and
1525 normal cases images. The experimental results showed that their system
achieved an accuracy near 99% with an AUC of 99.9%, a specificity of 99.2%, a
sensitivity of 99.3%, and a F1-score of 98.9%.
An inception model was proposed by Das et al., to screen Covid-19 from other
non-Covid cases like pneumonia, tuberculosis, and healthy cases [7]. The data
set included 162 Covid-19 images, 1583 pneumonia images and 342 tuberculosis
images. The proposed model achieved an AUC of 100% in classifying Covid-19
from a combined pneumonia and healthy cases data set. Similarly, it achieved
an AUC of 99% in classifying Covid-19 cases from a combined non-covid and
healthy diseases data set. This shows that it is easier to classify images when
there are fewer disease types in the data set. The fact that our data set contains
14 different illnesses alongside one for healthy cases is a reason for lower scores
when compared to those that contained only Covid-19 and healthy cases.
A CNN was implemented by Wang et al., to train the model over 13,975 chest
x-ray images to detect Covid-19 [8]. Their model was evaluated to a classification
accuracy of 98.9%.
A concatenated network was developed by Rahhimzadeh et al., with Xcep-
tion and ResNet50V2 models to classify Covid-19 chest x-ray images [9]. This
developed system contains 180 images of Covid-19, 6054 images of pneumonia,
and 8851 of healthy images. A total of 633 images were considered for the train-
ing phase, with an outcome of 99.56% of accuracy, 100% specificity, and 93.3%
sensitivity.
A Deep Bayes-SqueezeNet model is proposed by Ucar et al., to detect Covid-
19 images [10]. The data set includes three classes consisting of 76 Covid-19
images, 4290 pneumonia images, and 1583 normal images. The proposed model
achieved 98.3% accuracy for the Covid-19 class.
A deep neural network was designed by Khan et al., to automatically detect
and diagnose Covid-19 images [11]. The data set consists of 284 images of Covid-
19, 330 images of pneumonia-bacterial, 327 images of pneumonia-viral, and 310
images of healthy cases. Overall, the proposed system obtained 89.5% accuracy,
97% precision, and 100% recall for Covid-19 cases.
The system proposed by T. Ozturk at al., includes seven pre-trained models:
VGG19, MobileNetV2, InceptionV3, ResNetV2, DenseNet201, Xception, and
InceptionResNetV2 [12]. The data set consists of 25 images of Covid-19 and 25
images of non-Covid cases. Overall, VGG19 and DenseNet201 were selected as
4 S. Manjunath et M. Gurjar
the main model which achieved the highest values of 90% accuracy and 83%
precision.
A VGG16 deep transfer learning model was considered by Singh et al., to
detect Covid-19 from CT scans [13]. Principal Component Analysis was used for
feature extraction and trained for four different classes using ensemble method
and an SVM classifier resulting in 95.7% accuracy, 95.8% precision, and 95.3%
F1-score.
A review of the application of Machine-Learning and Deep-Learning in the di-
agnosis of Covid-19 through x-ray and CT images was conducted by Mohammad-
Rahimi et al [14]. 105 studies that used ML or DL methods to identify chest
images of Covid-19 patients were reviewed. All of these studies were conducted
on data sets that were much smaller than used in our study, in the range of 1000
- 10,000 images, with the largest study using a total of 60,000 images. Several
studies employed CNNs, of which, the highest accuracy, precision, sensitivity and
specificity were obtained by Chowdhury et al, 99.7%, 99.7%, 99.7% and 99.5%
respectively [15]. Butt et al., using a balanced data set of 618 images received
an AUC score of 99.6%, a sensitivity of 98.2% and a specificity of 92.2% [16].
Hassantabar et al., using a balanced data set of 682 images, received an accu-
racy of 93.2% a sensitivity of 96.1% and a specificity of 99.71% [17]. All of these
studies employing CNNs have relatively small balanced data sets, alongside a
low number of competing diseases: one or two diseases including Covid-19 (com-
pared to 15 in ours). Their higher results can be attributed to the fact that they
are undertaking an easier classification problem.
2.2 Techniques used for handling the imbalance problem:
During the pre-processing step, Lin et al., used two undersampling techniques
that involved clustering [18]. The first technique, used the cluster centres to rep-
resent the majority class, and the second technique, used the nearest neighbours
of the cluster centres. The experimental results revealed that the second strategy
outperformed the first.
Sun et al, proposed an ensemble method to convert the imbalanced data
sets into multiple balanced data sets and then built classifiers on them using a
classification algorithm [19]. The proposed model was compared with other sam-
pling methods, such as conventional sampling, a cost-sensitive learning method
and bagging and boosting based ensemble methods. The proposed method was
successful in resolving the imbalance to a greater degree.
Nanni et al., proposed a neighborhood balanced bagging technique to modify
samples according to the class distribution in their neighborhood using two ap-
proaches [20]. Firstly, they kept a large sample size through oversampling, and
secondly, they reduced the sample size using stronger undersampling.
3 Methodology
In this section, an overview of the approaches utilised in this study is given.
Detection of Covid-19 From an Imbalanced Chest X-ray Image Data Set 5
Table 1. Chest X-ray Image Data Set
Class No. Instances Total
Covid 3616 3616
Atelectasis - 11559
Cardiomegaly - 2776
Consolidation - 4667
Edema - 2303
Effusion - 13317
Emphysema - 2516
Fibrosis - 1686
Non-Covid 112,120
Infiltration - 19894
Mass - 5782
Nodule - 6331
Pleural thickening - 3385
Pneumonia - 1431
Pneumothorax – 5302
No findings - 60353
Total 115736
3.1 Data Set
For the identification and categorization of the Covid-19 illness, two types of
chest x-ray image data sets are used in our study. The Covid-19 images are
taken from the Covid-19 chest x-ray image data set which is freely available on
Kaggle [21]. The second set of x-ray images are taken from the National Institute
of Health (NIH) chest x-ray data set [22]. The chest x-ray is the most commonly
used imaging technique, due to its cost effectiveness, however, it poses more of a
challenge clinically than alternatives such as chest CT imaging. In reflection of
this, the model proposed in this study has been positioned to deliver clinically
meaningful diagnostics using a combination of chest x-ray imagery and publicly
available data sets that contain a large number of cases. The data set is relatively
large and complex owing to the many different disease types in the database.
From Table 1, it can be seen that just over half the observations have been
classified under the ’healthy’ or ‘No findings’ label. The data set allows for mul-
tiple label peer samples, however, a ‘No findings’ can only occur on its own. A
little over half of the observations in the general data set represent the absence
of disease and 3616 subjects had Covid-19. This amalgamated data set can be
found in a Github repository [23].
Each picture sample is accompanied by numerous fields in each of the two
data sets. These variables contain patient information such as gender, ID and
age, as well as information on the images perspective (PA or AP). The number
of images per class is presented in Table 1.
6 S. Manjunath et M. Gurjar
3.2 Image Pre-processing
The x-ray images used were taken at different orientations, positions, there was
some variation in size. The following pre-processing steps were taken: resize the
images to equal length and width; Gaussian Blur was used to remove the noise
from the sample image; smooth the edges of the images to allow easy detection;
segmentation was applied (histogram equalization), to normalize the brightness
and contrast.
3.3 Resampling Methods
In general, for machine learning or deep learning imbalanced learning problems
are dealt with by using resampling [4]. Resampling decouples the imbalance
problem from classification techniques allowing users to employ any standard
algorithm after the resampling pre-processing phase has been completed.
In this section, we present the different resampling methods for handling
imbalance in image data classification.
Oversampling Oversampling is a strategy that adds more samples to an imbal-
ance class. Many studies have used this strategy to solve classification challenges
for extremely imbalanced, image data sets.
There are many oversampling methods available which are designed as varia-
tions of the oversampling technique SMOTE such as: Borderline-SMOTE, Syn-
thetic Sampling (ADASYN), SVM-SMOTE [18]. SMOTE selects examples from
the feature space by drawing the line between the examples and by generating
samples along that line.
Instead of smaller particular decision areas, synthetic methods cause a clas-
sifier to produce bigger and less specific decision regions [20]. Thus, positive
examples are learned in more general regions rather than being impacted by
negative instances around the sample. SMOTE, however, runs into the problem
of over-generalization. It ignores the majority class and arbitrarily generalizes a
minority class region [19]. In the event of a highly skewed class distribution this
results in a higher likelihood of class mixing.
Undersampling Random undersampling in the training data set removes cer-
tain data points from the majority class. This results in a decreased number of
instances in the majority class of the training data set. This method is repeated
until each class has the same number of examples. This approach is suitable for
dealing with the imbalance when there are a sufficient number of samples in the
minority class that holds all the necessary information.
Undersampling approaches, may, on the other hand, result in a loss of infor-
mation, when data from the majority class are removed, leading to under-fitting.
No method exists to recognize or preserve information-rich instances from the
majority class, since examples are eliminated at random [18].
Detection of Covid-19 From an Imbalanced Chest X-ray Image Data Set 7
There are different approaches available making use of undersampling e.g.
TomekLinks. These approaches are used to solve problems with imbalanced data
sets; however, they are confined to moderate or low skewed data sets and are
not recommended for highly imbalanced data sets [20].
In order to avoid a low level of recall due to undersampling, this study applied
a combination of oversampling and undersampling, similar to the work by Sun
et al [19]. This process oversamples the minority class, and then undersamples
the majority class by 70%. It was found that making the ratio 70% minimised
the recall loss while reducing the imbalance problem.
3.4 Classification Models
Convolutional Neural Network In this study, a convolutional neural network
architecture was constructed to detect Covid-19 from chest x-ray images. The
network has 14 layers: six convolutional layers, four pooling layers, four dropout
layers, one fully connected layer, and one output layer. Each convolution block
is made up of two 2-D CNNs and one pooling layer, followed by a dropout layer
with a 25% dropout rate. The 5 x 5 kernel is used for feature extraction with
an ReLU activation function at the convolutional layer. The 3 x 3 kernel is used
by the max pooling layer to reduce the size of the input image. Finally, after
extracting the features, it is connected to fully connected layers for making the
prediction.
Transfer Learning Transfer learning is a deep learning approach that involves
training a neural network model on a problem that is similar to the one being
solved [4]. The learned model’s layers are then utilized in a new model that is
trained on the problem of interest.
In addition to the developed CNN model, four pre-trained models for classi-
fication have been used with applied transfer learning: ResNet50, NasNetLarge,
Xception and InceptionV3. ResNet50 is made up of convolutional layers with
an average pooling layer at the end. It uses residual learning as a building com-
ponent. CIFAR10 and ImageNet are two types of architectures that are part of
NasNetLarge [4]. The SoftMax function is implemented in the CIFAR10 architec-
ture, which includes N number of regular cells and one reduction cell repeating
after each other. ImageNet has the same architecture as CIFAR10, with two
strides of convolutional layers with a 3x3 kernel size at the start, followed by
two reduction cells [4]. In Xception there is one entry flow, eight intermediate
flows, and one exit flow [14]. Convolutional and max-pooling layers make up the
entry flow with a ReLU serving as the activation function.
Convolutional layers with a ReLU activation function are used in the in-
termediate flow on their own. Convolutional, max pooling and global average
pooling layers, with a ReLU activation function, make up the exit flow at the
end of the architecture [4]. The InceptionV3 model’s fundamental architecture
comprises convolutional and pooling layers, as well as three inception designs.
For the final layer, logistic and SoftMax functions are used.
8 S. Manjunath et M. Gurjar
To implement this structure, models were derived from the Python Keras
package and initialized with a shape of (80,80,1). The transfer learning models
are capable of categorizing a maximum of 1000 classes, however, for our purposes
all that was needed was two. Therefore the transfer learning models were modi-
fied by substituting the fully connected layer with one that has only outputs.
3.5 Evaluation Measures
The CNNs performances were assessed by using the following metrics: sensitivity,
recall, precision and F1-Score (F-Measure). All of these metrics may be evalu-
ated upon testing the model. The different outcomes for recall and precision is
discussed below [4]:
– High Precision & High Recall: the model performs well with the classification.
– High Precision & Low Recall: the model is unable to correctly categorize the
data points of a certain class or may overfit them.
– Low Precision & High Recall: the model correctly identifies data points from
a certain class, but incorrectly labels a large number of data points from
other classes.
– Low Precision & Low Recall: the model performed poorly in handling the
classification.
The Receiver Operating Characteristic (ROC) curve expresses the perfor-
mance of the classification CNN model used in this study. A graphical approach
is used that works at all classification thresholds. It is able to achieve this by
graphing the True Positive Rate and False Positive Rate. The metric gives a
summary of the performance of a classifier over all possible thresholds. Similar
to the ROC is the Area Under the ROC curve (AUC) which examines the entire
two-dimensional area underneath the entire ROC curve covering from (0,0) to
(1,1). This metric is effective at checking the proper or wellness and the quality
of our model’s prediction performance [3].
4 Results
The CNN and pre-trained models developed in this study, receive pre-processed
grayscale images as their input and binary classification is performed to distin-
guish between Covid and non-Covid cases. Meanwhile, the model is trained for
100 epochs and a batch size of 64 for each experiment.
To handle the imbalance problem, we tested the proposed model with a
combination of oversampling and undersampling as described in Section 3.3. The
new population of data was used for training the model. The obtained results
are displayed in Table 2. The CNN model that uses mixed sampling outperforms
the other combination of algorithms tested.
In machine learning, learning curves are a common diagnostic tool for al-
gorithms that learn progressively from a training data set. After each update
during training, the model can be assessed on the training data set and a hold
Detection of Covid-19 From an Imbalanced Chest X-ray Image Data Set 9
Table 2. Results obtained for proposed models when mixed sampling was done with
Random oversampling and Undersampling
Random Oversampling & Undersampling
Models F1-score(%) Precision(%) Recall (%) AUC(%)
CNN 93 90 91 94.5
ResNet50 74 68 86 68
InceptionV3 69 64 67 63
Xception 71 67 65 66
NasNetLarge 77 73 69 72
SMOTE and Random Undersampling
Models F1-score(%) Precision(%) Recall (%) AUC(%)
CNN 90 86 88 89
ResNet50 69 67 65 67
InceptionV3 63 61 62 61
Xception 65 64 64 65
NasNetLarge 71 70 68 70.5
out validation data set. Graphs of the measured performance can be generated
to display the learning curves. In this study, the training performances for both
SMOTE and Random Oversampling (ROS) were tested. A comparison of a learn-
ing curve is shown in Fig.4. Looking at these graphs, mixed sampling with ROS
performs well and there is no sign of overfitting or underfitting. On the other
hand, the mixed sampling with SMOTE graph, suggests that the model often
fluctuates at different epochs.
The classification evaluation metrics are summarised in the confusion ma-
trix which results in the True positive, True Negative, False Positive and False
Negative shown in Fig. 2.
The CNN model outperforms mixed sampling with random oversampling
approach by achieving 90% precision, 91% recall (this avoids unnecessary false
alarms) and 93% F1-score which measures the misclassified cases.
In the second phase of testing, pre-trained models were trained on different
combinations of resampling techniques. The results from this analysis were not
very significant. The large number of classes and continual overfitting of the
models were the main factors contributing to these uninformative results.
5 Conclusion and Future Work
Medical image analysis and feature extraction, which are used to diagnose a
wide range of chest illnesses, have benefited greatly from deep learning. CNN
architectures are well-known for their capacity to learn and predict mid-level
and high-level visual representations.
In this paper, a combination of over and under sampling methods has been
used to address the problem of class imbalance on the largest data set used for
x-ray detection of Covid-19. Image pre-processing was performed as described
10 S. Manjunath et M. Gurjar
Fig. 1. Accuracy and loss over 100 epochs of training and validation data set for ROS
and SMOTE of CNN model
Fig. 2. Confusion matrix for both ROS and SMOTE for CNN model
Detection of Covid-19 From an Imbalanced Chest X-ray Image Data Set 11
in Section 3.3 to remove unnecessary details and each model was trained with
all the combinations of the resampled data. The proposed modelling analysis
outperformed many of the existing methods.
The data set used here, was comparatively large relative to other research
conducted into x-ray classification of Covid-19. The presence of many different
illnesses in the data set poses a higher challenge in classification compared to
those works that used fewer disease types. Satisfactory results of F1-score, preci-
sion and recall have been obtained when compared to those obtained by previous
works. For the purposes of reproducibility and experimental validation, both the
data set and the code have been made publicly available.
Due to the exponential growth in Covid-19 reported cases and treatments
throughout the world, the content of Covid-19 databases are continuously grow-
ing by significant amounts. As a result, future research could focus on improving
the architecture of the deep-learning model proposed in this study, as well as
testing its robustness on even larger data sets.
In future work, the study could be expanded to explore the use of One-Class-
Classification algorithms used for imbalanced data sets, in conjunction with the
techniques that have been proposed in this paper.
References
1. Wang, W., Xu, Y., Gao, R., Lu, R., Han, K., Wu, G., Tan, W.: Detection of SARS-
CoV-2 in different types of clinical specimens. Jama, 323(18), pp.1843-1844., (2020)
2. Yang, T., Wang, YC., Shen, CF., Cheng, CM.: Point-of-care RNA-based diagnostic
device for COVID-19. Diagnostics, 165, (2020)
3. Oyelade, O.N., Ezugwu, A.E.: Deep Learning Model for Improving the Character-
ization of Coronavirus on Chest X-ray Images Using CNN. medRxiv, (2020)
4. Albahli, S., Yar, G.N.A.H.: Fast and Accurate Detection of Covid-19 Along With
14 Other Chest Pathologies Using a Multi-Level Classification: Algorithm Develop-
ment and Validation Study. Journal of Medical Internet Research, 23(2), p.e23693.,
(2021)
5. Loey, M., Smarandache, F., M Khalifa, N.E.: Within the Lack of Chest Covid-
19 X-ray Dataset: A Novel Detection Model Based on GAN and Deep Transfer
Learning. Symmetry, 12(4), p.651., (2020)
6. Islam, M.Z., Islam, M.M., Asraf, A.: A combined deep CNN-LSTM network for
the detection of novel coronavirus (COVID-19) using X-ray images. Informatics in
Medicine Unlocked, 20, p.100412., (2020)
7. Das, D., Santosh, K.C. and Pal, U.: Truncated inception net: COVID-19 outbreak
screening using chest X-rays. Physical and engineering sciences in medicine, 43(3),
pp.915-925., (2020)
8. Wang, L., Lin, Z.Q., Wong, A.: Covid-net: A tailored deep convolutional neural
network design for detection of covid-19 cases from chest x-ray images. Scientific
Reports, 10(1), pp.1-12., (2020)
9. Rahimzadeh, M., Attar, A.: A modified deep convolutional neural network for
detecting COVID-19 and pneumonia from chest X-ray images based on the con-
catenation of Xception and ResNet50V2. Informatics in Medicine Unlocked, 19,
p.100360., (2020)
12 S. Manjunath et M. Gurjar
10. Ucar, F., Korkmaz, D.: COVIDiagnosis-Net: Deep Bayes-SqueezeNet based diag-
nosis of the coronavirus disease 2019 (COVID-19) from X-ray images. Medical
Hypotheses, 140, p.109761., (2020)
11. Khan, A.I., Shah, J.L., Bhat, M.M.: CoroNet: A deep neural network for detec-
tion and diagnosis of COVID-19 from chest x-ray images. Computer Methods and
Programs in Biomedicine, 196, p.105581., (2020)
12. Ozturk, T., Talo, M., Yildirim, E.A., Baloglu, U.B., Yildirim, O., Acharya, U.R.:
Automated detection of COVID-19 cases using deep neural networks with X-ray
images. Computers in biology and medicine, 121, p.103792., (2020)
13. Singh, M., Bansal, S., Ahuja, S., Dubey, R.K., Panigrahi, B.K., Dey, N.: Transfer
learning-based ensemble support vector machine model for automated COVID-
19 detection using lung computerized tomography scan data. Medical biological
engineering computing, 59(4), pp.825-839., (2020)
14. H. Mohammad-Rahimi, M. Nadimi, A. Gjalyanchi-Langeroudi, M. Taheri, S.
Ghafouri-Fard: Application of Machine Learning in Diagnosis of Covid-19 Through
X-Ray and CT Images: A Scoping Review. Frontiers in cardiovascular medicine,
8, p.185., (2021)
15. M. Chowdhury, A. Khandakar, M. Kadir, Z. Bin Mahbub, K. Islam, M.Khan, A.
Iqbal, M. Reaz, M. Islam: Can AI Help in Screening Viral and Covid-19 Pneumo-
nia?. IEEE Access, 8, pp.132665-132676., (2020)
16. Butt, C., Gill, J., Chun, D., Babu, B. A.: Deep learning system to screen coron-
avirus disease pneumonia. Applied Intelligence, (2020)
17. S. Hassantabar, M. Ahmadi, A. Sharifi.: Diagnosis and detection of infected tissue
of Covid-19 patients based on lung x-ray image using convolutional neural network
approaches. Chaos, Solitons Fractals, 140, p.110170., (2020)
18. Lin, W.C., Tsai, C.F., Hu, Y.H., Jhang, J.S.: Clustering-based undersampling in
class-imbalanced data. Information Sciences, 409, pp.17-26., (2017)
19. Sun, Z., Song, Q., Zhu, X., Sun, H., Xu, B., Zhou, Y.: A novel ensemble method
for classifying imbalanced data. Pattern Recognition. 48(5), pp.1623-1637., (2015)
20. Nanni, L., Fantozzi, C., Lazzarini, N.: Coupling different methods for overcoming
the class imbalance problem. Neurocomputing, 158, pp.48-61., (2015)
21. Chowdhury, M.E., Rahman, T., Khandakar, A., Mazhar, R., Kadir, M.A., Mahbub,
Z.B., Islam, K.R., Khan, M.S., Iqbal, A., Al Emadi, N., Reaz, M.B.I.: Can AI help
in screening viral and COVID-19 pneumonia?. IEEE Access, 8, pp.132665-132676.,
(2020)
22. Wang, X., Peng, Y., Lu, L., Lu, Z., Bagheri, M., Summers, R.: Hospital-scale
Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and
Localization of Common Thorax Diseases. In IEEE CVPR (Vol. 7)., (2019)
23. Github, https://github.com/sharath30/Covid. Last accessed 15 Oct 2021