2024 International Conference on Advances in Computing, Communication,
Electrical, and Smart Systems (iCACCESS), 8-9 March, Dhaka, Bangladesh
Pneumonia Detection from Chest X-ray Images
using Convolutional Neural Network
Md. Zakir Hossain
Department of CSE
Green University of Bangladesh
Dhaka, Bangladesh
Email: mdzakir2551@gmail.com
Khalid Ibne Mostafa
Department of CSE
Green University of Bangladesh
Dhaka, Bangladesh
Email: khaledmd26937@gmail.com
Md. Mostafizur Rahman
Department of CSE
Green University of Bangladesh
Dhaka, Bangladesh
Email: mostafizurrahman0202@gmail.com
Md. Solaiman Mia
Department of CSE
Green University of Bangladesh
Dhaka, Bangladesh
Email: solaiman@cse.green.edu.bd
Abstract—The fact that Pneumonia ranks among the world’s
most common causes of mortality, precise diagnosis methods
are absolutely necessary. Although chest X-rays are a quick
way to identify Pneumonia, they can be challenging to interpret
due to the similarities between Pneumonia and other lung
conditions. This research presents a computer-aided approach to
Pneumonia diagnosis that uses chest x-rays to enhance diagnostic
decision-making. Moreover, situations such as the coronavirus
pandemic, where widespread lockdowns are implemented and
human contact poses significant risks, highlight the importance
of computer-aided diagnosis like this. Consequently, a technique
for the quick and automatic diagnosis of Pneumonia is presented
in this work. Based on chest X-ray pictures, a deep learning-
based architecture called “ESPD” is suggested for the automatic
diagnosis of Pneumonia. A benchmark dataset comprising 5,856
chest X-ray pictures was utilized for the suggested deep-learning
network’s testing, evaluation, and training. The total accuracy
of the suggested model was found to be 98.24%, consisting of
0.98 F1-Score, 0.98 precision, 0.98 recall, and 0.97 specificity. In
comparison to previous methods in the literature, the suggested
method was found to be quicker and less computationally
expensive, and it also produced accuracy that showed promise.
Keywords—Pneumonia, Chest X-ray, Data Augmentation, Con-
volutional Neural Network.
I. I NTRODUCTION
Chronic infections of the lungs like Pneumonia affect an
excessive number of young children around the globe every
year. Pneumonia was the primary reason for among children
under the tender age of five between 1990 and 2019 [1].
In areas with poor healthcare services, the most common
cause of death for children and young adults is Pneumonia.
The lack of a reliable healthcare system in many developing
and middle-income countries compounds the problem. The
elderly and those with prior health risks are particularly at
risk from this condition. Rapid breathing, discomfort in the
chest area, prolonged coughing, initial cold-like symptoms and
difficulties maintaining normal breathing patterns are all signs
of Pneumonia, as defined by the World Health Organization
(WHO) [2].
Sustainability integrates the utilization of Convolutional
Neural Network (CNN) for identifying cases of Pneumonia,
combining both environmentally conscious procedures with
medical evaluation [3]. The utilization of CNNs for imaging
purposes boosts Pneumonia determination, resulting in rapid
treatment. Industry 5.0 ensures environmentally friendly oper-
ations and environmentally compatible hardware options. This
collaboration serves equally leading-edge medical services and
sustainable practices research and development initiatives [3].
The laboratory preferred method for Pneumonia is ra-
diography (chest X-rays). Notwithstanding this, psychiatrists
continue analyzing images individually, in spite of being
constrained due to variables like exhaustion and their lack of
experience. It necessitates quite a bit of effort and financial
resources to turn into an appropriately trained radiology [4].
The rural parts of a nation with low revenues additionally
experience a lack of radiologists. With the resemblance to
other illnesses, evaluating Pneumonia via an X-ray of the chest
may prove challenging even to highly qualified radiologists.
If a doctor gives the inaccurate diagnosis, for instance, the
individual in question might miss the medical treatment they
require and can die as the consequence. Thus, Pneumonia
could be problematic to identify properly [5].
For all of these factors, developing of Computer Aided
Diagnosis (CAD) to assist in determining the presence of
Pneumonia is of the greatest possible significance [6]. In recent
years, CAD methods based on Deep Learning (DL) are gaining
acceptance for use in the field of medicine. CNN, a type
of DL-based CAD system inspired by the visual cortex of
people, are able to learn from a wide range of information,
particularly data that has been appraised by psychiatrists
[7]. This paper introduces a deep learning based architecture
called ESPD (Early Stage Pneumonia Detection) specifically
designed for the automatic diagnosis of Pneumonia from chest
X-ray images. Our proposed model achieves a high accuracy
of 98.24% on a benchmark dataset consisting of 5,856 chest
X-ray images.
979-8-3503-5028-9/24/$31.00 © 2024 IEEE