Pneumonia Detection in X-Ray Chest Images Based on Convolutional

Neural Networks and Data Augmentation Methods

Samia Dardouri

1,2 a

Department of Computer Science, College of Computing and Information Technology, Shaqra University, Saudi Arabia

InnoV'COM Laboratory-Sup'Com, University of Carthage, Tunisia

Keywords: CNN, Feature Extraction, Pneumonia Infection, Image Data Augmentation, Deep Learning, Adam Optimizer,

Early Detection.

Abstract: Pneumonia, a widespread lung ailment, stands as a leading global cause of mortality, particularly affecting

vulnerable demographics such as children under five, the elderly, and individuals with underlying health

conditions. Accounting for a significant portion of childhood fatalities, at 18%, pneumonia remains a critical

health concern. Despite advancements in imaging diagnostic methods, chest radiographs remain pivotal due

to their cost-effectiveness and rapid results. The proposed model, trained on data sourced from a readily

available Kaggle database, consists of two primary stages: image preprocessing and feature extraction/image

classification. Utilizing a CNN model, the framework achieves remarkable performance metrics, with

precision, recall, F1-score, and accuracy reaching 93%, 96%, 94%, and 96%, respectively. These results

underscore the CNN model's effectiveness in pneumonia detection, showcasing superior consistency and

accuracy compared to other pretrained deep learning models.

1 INTRODUCTION

Pneumonia is a prevalent and potentially life-

threatening respiratory infection, particularly

affecting vulnerable populations such as children, the

elderly, and immunocompromised individuals. Early

detection of pneumonia is crucial for timely

intervention and treatment to prevent complications

and improve patient outcomes. While numerous

imaging diagnostic methods exist, many are

prohibitively expensive and inaccessible to large

segments of the population, especially in low-

resource regions Asnake, N.W (2024). Additionally,

the shortage of radiology experts in these areas and

long waiting times for diagnoses exacerbate the

severity of the disease and contribute to increased

mortality rates. Diagnostic radiography, although

cost-effective and rapid, may lead to

misinterpretations due to visualized opacities.

To address these challenges, recent studies have

explored machine learning techniques, particularly

deep learning models like convolutional neural

networks (CNNs), to aid in pneumonia diagnosis

using high-resolution imaging modalities such as

https://orcid.org/ 0000-0002-1376-9607

computed tomography (CT) scans. One such study

developed a deep learning architecture tailored for

diagnosing severe pneumonia cases from chest X-

rays. Leveraging a dataset from the Radiological

Society of North America, this study focused on

specialized zones within the chest X-rays for

improved diagnostic accuracy.

In recent years, the healthcare landscape has

witnessed the emergence of various technologies

such as genomics and imaging, which have brought

forth vast and intricate datasets in Asnake, N.W.,

Salau (2024). While chest X-ray images remain a

primary diagnostic tool for pneumonia, they can pose

challenges due to their nuanced nature, sometimes

leading to misclassifications by expert radiologists

and subsequent incorrect treatments (Lamia A, Fawaz

A (2022)). This underscores the need for an automatic

and intelligent model to aid radiologists in accurately

diagnosing different types of pneumonia from chest

X-ray images Goyal, S., Singh, R (2023).

Deep learning, a subset of machine learning

inspired by the brain's structure and function, has

emerged as a powerful tool in medical image analysis

Kareem, A., Liu, H (2022). These algorithms excel at

Dardouri, S.

Pneumonia Detection in X-Ray Chest Images Based on Convolutional Neural Networks and Data Augmentation Methods.

DOI: 10.5220/0013147300003938

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 11th International Conference on Information and Communication Technologies for Ageing Well and e-Health (ICT4AWE 2025), pages 165-172

ISBN: 978-989-758-743-6; ISSN: 2184-4984

165

quantifying, identifying, and classifying patterns

within medical images by learning features directly

from data, eliminating the need for manual feature

design based on domain-specific knowledge.

Convolutional neural networks (CNNs) are a

prominent example of deep learning models utilized

in this context. These layers specialize in processing

images and extracting low-level features, such as

edges, while efficiently capturing temporal and

spatial dependencies with the aid of filters. Unlike

traditional feed-forward layers, CNNs significantly

reduce computational complexity by sharing weights

and utilizing fewer parameters. As a result, CNNs

offer an effective approach for medical practitioners

to diagnose and classify specific medical conditions

with accuracy (Z. Li et al 2019 ; M. K. Gourisaria

2023). The structure of the paper is as follows:

Section 2 provides a detailed review of related works,

summarizing relevant research on the topic. Section 3

describes the dataset utilized in this study. Section 4

outlines the proposed methodology implemented in

the research. Section 5 examines the results and

compares them with findings from recent studies.

Finally, Section 6 concludes the paper by

summarizing key insights and proposing directions

for future research.

2 RELATED WORKS

In the field of disease detection, numerous

researchers have been actively engaged in developing

automated detection models. Deep learning

techniques have emerged as valuable tools for

enhancing productivity, especially in computer-

assisted diagnosis technologies, notably within

medical imaging, image classification, and image

restoration Venkateswara Reddy. (2022). Author in

Shagun Sharma. (2023) proposed deep learning (DL)

model comprises several stages: data collection,

preprocessing, feature extraction, training, testing,

classification, and pneumonia prediction. During data

preprocessing, the data is balanced and normalized,

ensuring it falls within a normalized range of [0-255].

Subsequently, the normalized data is inputted into the

VGG16 model for feature extraction Liu, Y (2023).

This step involves extracting pertinent features from

the images, facilitating the classification and

prediction process. With its 16 layers encompassing

input, convolution, pooling, dense, and output layers,

VGG16 enables comprehensive feature extraction.

The significant challenges faced in pneumonia

detection include the large number of patients and the

shortage of medical experts and supporting staff. The

development of deep learning-based methods for

early detection of pneumonia has garnered significant

attention in recent years due to their potential to

improve diagnostic accuracy and efficiency. By

leveraging advanced computational techniques and

large datasets of annotated medical images,

researchers have made significant strides in

developing deep learning models capable of detecting

pneumonia infections at an early stage Shadi A.

(2022).

In the study discussed in Lamia A. (2022),

pneumonia emerges as a rapidly spreading disease,

posing significant risks to individuals' health and

well-being. Biomedical diagnosis of pneumonia

typically involves a range of diagnostic tools and the

assessment of various clinical features. However,

limitations in expert availability and tool accessibility

hinder these efforts. To address this challenge, the

researchers are developing a mobile application

employing deep learning techniques to classify

pneumonia cases. The aim is to create a prototype

mobile app capable of detecting pneumonia using

neural networks. Utilizing high-level tools like Create

ML simplifies the process by eliminating

complexities such as determining neural network

layers, initializing model parameters, or selecting

algorithms. This approach enables broader

accessibility to the model, allowing users to access it

via a mobile application. With a dataset comprising

over 5,000 real images, an image classification model

is trained using Create ML, a tool that offers a user-

friendly graphical interface, requiring no specialized

knowledge for operation.

In their study Khalaf Alshamrani. (2022), the

authors optimized a model using data augmentation

techniques, resulting in slightly better precision

compared to the original model. They utilized this

improved model to develop a web application capable

of processing images and providing predictions to

users. The classification model they developed

achieved a prediction accuracy of 78%. The authors

noted that precision could be further enhanced by

adjusting parameters such as the number of epochs.

Their research aimed to showcase the potential of

artificial intelligence in creating deep-learning

models to aid healthcare professionals in early

pneumonia detection, emphasizing the importance of

such technology in public health initiatives.

In Dalya S. (2022), a deep learning model is

introduced for the detection of pneumonia disease

from chest X-ray images. It is noted that the number

of layers does not consistently lead to improved

accuracy, and increasing the number of layers in

neural networks may sometimes result in decreased

ICT4AWE 2025 - 11th International Conference on Information and Communication Technologies for Ageing Well and e-Health

166

performance. During the construction of the CNN

model, an optimal number of layers was determined,

which resulted in the highest accuracy achieved.

In Rajasenbagam, T., (2023), researchers

introduced a Deep Convolutional Neural Network

(CNN) aimed at detecting pneumonia infection in

lung tissues using chest X-ray imagery. The Deep

CNN models were trained on a Pneumonia Chest X-

ray Dataset consisting of 12,000 images depicting

both infected and uninfected chest X-rays. This

dataset underwent preprocessing and was curated

from the Chest X-ray8 dataset. Through the

application of a Content-based image retrieval

technique, images within the dataset were annotated

with metadata and additional content information.

Data augmentation techniques were then employed to

expand the image count in each class Farhan, A.M.Q

(2023), utilizing basic manipulation methods and the

Deep Convolutional Generative Adversarial Network

(DCGAN). The VGG19 network was employed in

the development of the proposed Deep CNN model.

Notably, this model achieved a classification

accuracy of 99.34% when tested on unseen chest X-

ray images D. S. V. Kancherla (2023).

Many studies have introduced methodologies

aimed at addressing the challenge of class imbalance.

One such approach involves leveraging Generative

Adversarial Networks (GANs), specifically a fusion

of Deep Convolutional Generative Adversarial

Network (DCGAN) and Wasserstein GAN with

gradient penalty (WGAN-GP), to augment the

minority class "Pneumonia." Concurrently, Random

Under-Sampling (RUS) techniques are employed on

the majority class "No Findings" to mitigate the

effects of class imbalance (Shorten, C.2019 and

Schaudt, D 2023). Various researchers have utilized

AI and CNN-based techniques for pneumonia

detection, as outlined in Table 1.

This study aims to introduce an efficient deep

learning framework tailored for pneumonia detection

using chest X-ray images, achieving a harmonious

balance between accuracy and complexity while

offering a cost-effective solution for medical and

radiology professionals. The outlined objectives are

as follows:

• Utilizing a CNN model to detect pneumonia

from chest X-ray images, serving as a feature

extraction and classification scheme.

• Exploring and evaluating the performance of

CNN and other deep learning models in

accurately classifying pneumonia cases.

• Developing a versatile model capable of

discerning between normal and abnormal

(pneumonia) chest X-ray images.

Table 1: Comparison of the results with some state of the

art methods.

Study Dataset Method

ccuracy

Rate

Lamia A.

2022

The dataset of more than

5,000 real images

Multilayer

Perceptron

(MLP),

Random

forest,

Sequential

Minimal

Optimization

(SMO)

84%

Shagun

Sharma.

2023

https://www.kaggle.com/da

tasets/prashant268/chest-

xray-covid19-pneumonia

Vgg16 92.15%

Jain DK.

2022

“Curated Dataset for

COVID-19 Posterior-

Anterior Chest Radiography

Images (X-Rays)

Vgg16 94%

Vgg19 95%

Xception 96%

Goyal, S.

2023

Covid-19 Radiography

Database (C19RD)

collected from Kaggle

(https://www.kaggle.com/ta

wsifurrahman/covid19-

radiography-database)

F-RNN-

LSTM

95.04%

Fatma

Taher.

2022

CXR images were produced

at the Rashid Hospital

Radiology Department in

Dubai in the United Arab

Emirates

CNN

94%

Proposed

model

lung disease dataset

collected from Kaggle

(https://www.kaggle.com/

paultimothymooney/chest

-xray-pneumonia)

CNN+Adam

optimizer

96%

3 DATASET DESCRIPTION

The dataset, sourced from Kaggle, is organized into

two main directories: "train" and "test." Each

directory contains subdirectories, one containing X-

ray radiographs of pneumonia cases and the other

containing radiographs of normal lungs.

Specifically, 5,856 anteroposterior CXR images

from pediatric patients aged 1 to 5 years were selected

for analysis. Two labels, "pneumonia" and "normal,"

were assigned to categorize the images accordingly.

Following an adjustment and consolidation of initial

data classifications, the entire image dataset was split

into 70% for training and 30% for testing purposes.

This allocation was made to ensure a

comprehensive evaluation of the system's

performance. Consequently, 5,216 X-ray images

were allocated for training, while 640 radiographs

were reserved for testing the system's efficacy. These

images are chest X-rays (anterior-posterior) obtained

from retrospective cohorts of pediatric patients aged

one to five years old at Guangzhou Women and

Pneumonia Detection in X-Ray Chest Images Based on Convolutional Neural Networks and Data Augmentation Methods

167

Children’s Medical Center. The chest X-ray imaging

was conducted as part of the routine clinical care of

the patients.

Figure 1: Samples of the dataset.

Before inclusion in the dataset, all chest

radiographs underwent quality control screening to

eliminate any low-quality or unreadable scans.

Subsequently, the diagnoses for the images were

assessed by two expert physicians to ensure accuracy

before being used for training the AI system.

Additionally, a third expert verified the evaluation set

to address any potential grading errors, further

enhancing the reliability of the dataset.

4 PROPOSED METHODOLOGY

The proposed deep learning framework has

undergone multiple constructions and training

sessions, exploring various parameters to select

optimal hyperparameters and achieve a balanced

performance architecture. Broadly, it comprises two

primary stages. The initial stage involves several

images preprocessing steps, including image resizing

to obtain a standardized size and rescaling pixel

values to fall within the [0,1] interval. Subsequently,

the second stage focuses on feature extraction and

image classification using the proposed

Convolutional Neural Network (CNN) models

(figure2).

A CNN consists of multiple layers, including

convolutional layers, pooling layers, and fully

connected layers, that work together to automatically

learn spatial hierarchies of features from input images.

The convolutional layers apply filters (or kernels) to

the images to detect local patterns such as edges,

textures, and shapes. Pooling layers reduce the spatial

dimensions of the data, retaining important features

while improving computational efficiency. Finally,

the fully connected layers at the end of the network

make predictions based on the features learned by the

convolutional and pooling layers.

Figure 2: CNN structure (Sun, Shuo & Sun. 2022).

The feature extraction stage constitutes the second

component of the CNN architecture, comprising three

blocks, each containing a convolution layer,

maximum pooling layer, and dropout layer. Within

the convolutional layer, input images are transformed

into matrix representations. The convolution

operation is applied between the input matrix and a

feature kernel of a specified dimension, resulting in a

feature map. This operation effectively reduces the

dimensions of the image, facilitating further

processing. Data augmentation methods [23-24]

prove beneficial in addressing the imbalance and

scarcity of data in certain classes when dealing with

limited and uneven datasets (Figure 3). This approach

proves particularly useful for achieving a

balance in

the number of images across different MRI classes

related to brain tumors and for expanding the overall

dataset. Various augmentation techniques, including

rotation, cropping, height and width adjustments,

filling operations, zooming, and horizontal rotation

brightening, are employed to augment images and

rectify class imbalances. Given the unbalanced nature

of our dataset, this augmentation technique is applied

to artificially increase the number of images for each

class, particularly those with fewer instances.

Figure 3: Imbalanced Data.

ICT4AWE 2025 - 11th International Conference on Information and Communication Technologies for Ageing Well and e-Health

168

Given the apparent class imbalance in the dataset,

with potentially fewer instances of the "Pneumonia"

class compared to the "Normal" class, a strategy to

counter this issue involves leveraging data

augmentation techniques. By employing data

augmentation, we aim to augment the training dataset

by generating synthetic examples, thereby increasing

the number of instances available for training.

This approach not only addresses class imbalance

but also enhances the robustness and generalization

ability of the machine learning model. To mitigate the

risk of overfitting, expanding our dataset through

artificial means is crucial. This involves introducing

variations to the existing data via minor

transformations, thereby increasing its size.

Techniques that manipulate training data while

preserving their labels are known as data

augmentation methods . Common augmentations

include grayscale conversions, horizontal and vertical

flips, random cropping, colour adjustments,

translations, rotations, and more. By applying a

subset of these transformations to our training data,

we can substantially augment the number of

examples, leading to the creation of a highly robust

model.

For data augmentation, I implemented several

transformations to enhance the training dataset,

including randomly rotating some images by up to 30

degrees, zooming in or out by up to 20%, shifting

images horizontally by 10% of their width and

vertically by 10% of their height, and randomly

flipping images horizontally. These techniques were

applied to increase the diversity of the training data

and improve the model's ability to generalize. Once

our model is prepared, we proceed to fit the

augmented training dataset.

The imbalance in clinical

datasets, with a majority of abnormal cases and fewer

normal cases, could indeed lead to overfitting, as the

model might disproportionately favour the majority

class. To mitigate this, our study implemented data

augmentation techniques to increase the diversity and

representation of normal cases through

transformations like rotation, flipping, and scaling.

Additionally, we employed class balancing strategies,

such as adjusting class weights during training, to

penalize misclassification of the minority class more

heavily.

We validated the model using stratified cross-

validation to ensure an even class distribution across

folds and monitored evaluation metrics such as recall,

precision, and F1-score, which are sensitive to

imbalanced data. These approaches ensured the

development of a balanced and reliable model for

pneumonia detection. The proposed approach

consists of two main steps. Firstly, we introduce a

normalization method specifically designed for chest

X-rays, with the goal of removing unnecessary

components while retaining crucial information.

Following this, Deep Convolutional Neural Networks

(CNNs) are utilized, with a preference for using the

ADAM optimization function to build predictive

models using the normalized dataset.

ADAM

combines the benefits of the Adaptive Gradient

Algorithm (AdaGrad) and Root Mean Square

Propagation (RMSProp), computing adaptive

learning rates for each parameter to enhance training

efficiency and convergence Figure 4 provides an

overview of the proposed approach. The model,

consisting of 6,026,324 parameters, employs a multi-

branch convolutional architecture with three distinct

branches, each featuring different lengths and kernel

sizes to optimize feature extraction. Smaller kernels

specialize in detecting localized features such as

edges and textures, while larger kernels capture more

global patterns like shapes and contours.

By varying the branch depths, the model

combines shallow layers for basic feature recognition

with deeper layers that learn complex, high-level

abstractions. This design enables the model to process

input data at multiple scales, enriching its feature

representation and enhancing its ability to analyse

fine-grained details alongside broader patterns.

Figure 4: Proposed Architecture.

5 RESULTS AND DISCUSSION

For assessing the performance of the model, several

evaluation metrics are employed to gauge the

classification outcomes, particularly for pneumonia

Pneumonia Detection in X-Ray Chest Images Based on Convolutional Neural Networks and Data Augmentation Methods

169

classification from lung X-rays. The primary

evaluation metrics utilized include accuracy, recall

(sensitivity), and F1-score. These metrics are

calculated using the following equations:

Accuracy=





(1)

Precision=





.. (2)

Sensitivity(Recall)=





(3)

F1-score= 2/((1/𝑃𝑟𝑒𝑐𝑖𝑠𝑖on) + (1/Recall)) (4)

Here, TP denotes true positives, TN denotes true

negatives, FP denotes false positives, and FN denotes

false negatives. These metrics collectively provide a

comprehensive evaluation of the model's

performance in detection pneumonia cases from chest

X-ray images. Figure 5 illustrate view of the pneumonia

detection application. The initial step involves obtaining the

patient's information, including their name, gender, age,

phone number, history of hypertension, and the

neurologist's name. Once the user correctly fills out this

information, they are prompted to upload the lung X-ray

images. Upon submitting the chest X-Ray Images, the

application initiates the analysis using learning models to

determine the presence of a pneumonia infection.

Figure 5: Pneumonia Detection Application.

Final output of the proposed method using 12 numbers

of epoch has been shown in Figure 6. shows the model

accuracy and the corresponding loss with respect to the

number of epochs.

Our proposed model outperforms the previously

developed approaches demonstrating accuracy of

96% and the loss is 1%. Our experimental results

illustrate that the proposed CNN model exhibits

superior convergence compared to the ANN

approach, Random Forest classifier, Transfer

learning algorithms, and other CNN models. As

indicated in Table 2, our model attained the highest

accuracy rate of 96% and the best F1-Score of 94%,

along with a precision of 93%. The values attained by

our proposed CNN model stand out for their

Figure 6: The model accuracy and loss over the epochs

exceptional performance, surpassing those achieved

by the previously mentioned models. Table 2 presents

a comparison of pneumonia detection accuracy

between our proposed novel framework and state-of-

the-art models. While our proposed model achieved

accuracy that surpasses the state-of-the-art, it's

important to note that directly comparing accuracy

may not be entirely objective.

Table 2: Comparative study.

Study Method

Accuracy

Rate

Lamia A. 2022 Multilayer Perceptron

(MLP), Random

forest, Sequential

Minimal Optimization

(SMO)

84%

Shagun Sharma.

2023

Vgg16 92.15%

Jain DK. 2022 Vgg16 94%

Vgg19 95%

Xception 96%

Goyal, S. 2023 F-RNN-LSTM 95.04%

Fatma Taher.

2022

CNN 94%

Proposed model CNN+Adam

optimizer

96%

ICT4AWE 2025 - 11th International Conference on Information and Communication Technologies for Ageing Well and e-Health

170

6 CONCLUSIONS AND FUTURE

WORK

This study presents an automated method for

pneumonia detection using X ray scans, leveraging a

deep learning model for automated feature extraction

from the images. The main goal of this research was

to achieve improved classification performance with

faster learning rates compared to traditional deep

learning (DL) models. Despite the limited training

data available, experimental results demonstrate the

effectiveness of the proposed model. Its success can

be attributed to minimal preprocessing requirements

and the absence of handcrafted features, making it

suitable for diverse x ray classifications. Future

research aims to expand the classification to include

additional labels while enhancing accuracy. Future

work should aim to validate the proposed system

beyond Chest X-ray (CXR) images. It is imperative

to extend the validation to include other imaging

modalities such as computerized tomography (CT)

scans and Magnetic Resonance Imaging (MRI). This

expansion of validation will enhance the applicability

and robustness of the system across various medical

imaging techniques. Future work in pneumonia

detection using X-ray chest images could focus on the

exploration of more advanced architectures, such as

deeper or hybrid convolutional neural network

(CNN) models, which could improve detection

accuracy by capturing more complex features and

patterns. Additionally, the integration of transfer

learning from pre-trained models on large, diverse

datasets could significantly enhance performance,

particularly when labelled training data is scarce.

ACKNOWLEDGEMENTS

We would like to thank the Deanship of Scientific

Research at Shaqra University for supporting this

work.

REFERENCES

Lamia A, Fawaz A. Detection of Pneumonia Infection by

Using Deep Learning on a Mobile Platform. Comput

Intell Neurosci. 2022 Jul 30;2022:7925668. doi:

10.1155/2022/7925668. PMID: 35942467; PMCID:

PMC9356824.

Shagun Sharma, Kalpna Guleria, A Deep Learning based

model for the Detection of Pneumonia from Chest X-

Ray Images using VGG-16 and Neural Networks,

Procedia Computer Science, Volume 218, 2023.

Jain DK, Singh T, Saurabh P, Bisen D, Sahu N, Mishra J,

Rahman H. Deep Learning-Aided Automated

Pneumonia Detection and Classification Using CXR

Scans. Comput Intell Neurosci. 2022 Aug

4;2022:7474304. doi: 10.1155/2022/7474304. PMID:

35936981; PMCID: PMC9351538.

Asnake, N.W., Salau, A.O. & Ayalew, A.M. X-ray image-

based pneumonia detection and classification using

deep learning. Multimed Tools Appl (2024).

https://doi.org/10.1007/s11042-023-17965-4

Goyal, S., Singh, R. Detection and classification of lung

diseases for pneumonia and Covid-19 using machine

and deep learning techniques. J Ambient Intell Human

Comput 14, 3239–3259 (2023). https://doi.org/10.1007

/s12652-021-03464-7

Kareem, A., Liu, H. & Sant, P. Review on Pneumonia

Image Detection: A Machine Learning Approach.

Hum-Cent Intell Syst 2, 31–43 (2022). https://doi.org/

10.1007/s44230-022-00002-2

Z. Li et al., "PNet: An Efficient Network for Pneumonia

Detection," 2019 12th International Congress on Image

and Signal Processing, BioMedical Engineering and

Informatics (CISP-BMEI), Suzhou, China, 2019, pp. 1-

5, doi: 10.1109/CISP-BMEI48845.2019.8965660.

M. K. Gourisaria, V. Singh, R. Chatterjee, S. K. Panda, M.

R. Pradhan and B. Acharya, "PneuNetV1: A Deep

Neural Network for Classification of Pneumothorax

Using CXR Images," in IEEE Access, vol. 11, pp.

65028-65042, 2023, doi: 10.1109/ACCESS.2023.328

9842.

Venkateswara Reddy, E., Naveen Kumar, G.S., Siva Naga

Dhipti, G., Swathi, B. (2022). Diagnosis of Pneumonia

with Chest X-Ray Using Deep Neural Networks. In:

Reddy, V.S., Prasad, V.K., Mallikarjuna Rao, D.N.,

Satapathy, S.C. (eds) Intelligent Systems and

Sustainable Computing. Smart Innovation, Systems and

Technologies, vol 289. Springer.

Shadi A. Aljawarneh and Romesaa Al-Quraan.Pneumonia

Detection Using Enhanced Convolutional Neural

Network Model on Chest X-Ray Images. Big

Data.ahead of print http://doi.org/10.1089/big.2022.02

61.

D. S. V. Kancherla, P. Mannava, S. Tallapureddy, V.

Chintala, K. P and C. Iwendi, "Pneumothorax: Lung

Segmentation and Disease Classification Using Deep

Neural Networks," 2023 International Conference on

Self Sustainable Artificial Intelligence Systems

(ICSSAS), Erode, India, 2023, pp. 181-187, doi:

10.1109/ICSSAS57918.2023.10331853.

Farhan, A.M.Q., Yang, S. Automatic lung disease

classification from the chest X-ray images using hybrid

deep learning algorithm. Multimed Tools Appl 82,

38561–38587 (2023). https://doi.org/10.1007/s11042-

023-15047-z.

Liu, Y., Liang, P., Liang, K. et al. Automatic and efficient

pneumothorax segmentation from CT images using

EFA-Net with feature alignment function. Sci Rep 13,

15291 (2023). https://doi.org/10.1038/s41598-023-

42388-4

Pneumonia Detection in X-Ray Chest Images Based on Convolutional Neural Networks and Data Augmentation Methods

171

Chest X-Ray images (Kaggle) https://www.kaggle.com/

datasets/paultimothymooney/chest-xray-pneumonia

Khalaf Alshamrani, Hassan A. Alshamrani, Abdullah A.

Asiri, F. F. Alqahtani, Walid Theib Mohammad, Ali H.

Alshehri, "[Retracted] The Use of Chest Radiographs

and Machine Learning Model for the Rapid Detection

of Pneumonitis in Pediatric", BioMed Research

International, vol. 2022, Article ID 5260231, 11 pages,

2022.https://doi.org/10.1155/2022/5260231

Dalya S. Al-Dulaimi, Aseel Ghazi Mahmoud, Nadia

Moqbel Hassan, Ahmed Alkhayyat, Sayf A. Majeed,

"Development of Pneumonia Disease Detection Model

Based on Deep Learning Algorithm", Wireless

Communications and Mobile Computing, vol. 2022,

Article ID 2951168, 10 pages, 2022. https://doi.org/

10.1155/2022/2951168

Rajasenbagam, T., Jeyanthi, S. & Pandian, J.A. Detection

of pneumonia infection in lungs from chest X-ray

images using deep convolutional neural network and

content-based image retrieval techniques. J Ambient

Intell Human Comput (2021). https://doi.org/10.1007/

s12652-021-03075-2

Shorten, C., Khoshgoftaar, T.M. A survey on Image Data

Augmentation for Deep Learning. J Big Data 6, 60

(2019). https://doi.org/10.1186/s40537-019-0197-0

Schaudt, D., von Schwerin, R., Hafner, A. et al.

Augmentation strategies for an imbalanced learning

problem on a novel COVID-19 severity dataset. Sci

Rep 13, 18299 (2023). https://doi.org/10.1038/s41598-

023-45532-2https://doi.org/10.1155/2022/2656818

Sun, Shuo & Sun, Junzhong & Wang, Zongliang & Zhou,

Zhiyong & Cai, Wei. (2022). Prediction of Battery SOH

by CNN-BiLSTM Network Fused with Attention

Mechanism. Energies. 15. 4428. 10.3390/en15124428.

ICT4AWE 2025 - 11th International Conference on Information and Communication Technologies for Ageing Well and e-Health

172