Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Applying Medical Technologies for Diagnoising

Medical Images by Using Machine Learning
B.Nagaraju1 R.Srija2
Assistant Professor -IT, NRI Institute of Technology, UG Scholar, Dept. Of IT, NRI Institute of Technology,
A.P, India-521212 A.P, India-521212

M.Komala3 P.Dharani4
UG Scholar, Dept. Of IT, NRI Institute of Technology, UG Scholar, Dept. Of IT, NRI Institute of Technology,
A.P, India-521212 A.P, India-521212

Abstract:- Medical imaging is important in a variety of The ideas of cognition and information were the
clinical activities, including early detection, monitoring, foundation for the idea of deep learning algorithms. Deep
an opinion, and therapy evaluation of many medical learning often possesses two characteristics: (1) many
diseases. grasp medical image analysis in a computer processing layers that may learn unique data features
vision requires a solid grasp of the principles and through various degrees of generalization, and (2)
operations of artificial neural networks, as well as deep unsupervised or supervised learning of feature presentations
literacy. Deep Learning Approach (DLA) in medical on each layer. The possibilities of improved DLA in the
image processing is emerging as a rapidly increasing medical fields of MRI, Radiology, Cardiology, and
research subject. DLA has been widely utilised in Neurology have been emphasized in a growing number of
medical imaging to characterise the presence or absence recent review studies. supervised deep learning methods
of a complaint. The vast majority of DLA executions include recurrent neural networks (RNNS) and
focus on X-ray pictures, motorised tomography images, convolutional neural networks. Medical image processing
mammography images, and digital histopathology has also been studied using unsupervised learning methods
images. It presents a rigorous assessment of studies like Deep Belief Networks and Generative Adversarial
based on DLA for bracketing, discovery, and Networks (GAN'S). DLA can be used to identify
segmentation of medical pictures. This review directs the abnormalities and categorize certain types of diseases.
experimenters' assumptions.
II. TECHNOLOGIES USED
Keywords:- Artificial Neural Networks, Deep Literacy,
Deep Learning Approach (DLA), Motorized Tomography,
Mammography Images, Digital Histopathology Images.

I. INTRODUCTION

Medical image services, such as radiography,

colonoscopy, computerised tomography (CT),
mammography images (MG), ultrasound images, magnetic
resonance imaging (MRI), magnetic resonance angiography
(MRA), nuclear medicine imaging, positron emission
tomography (PET), and pathological tests, have seen an
increase in demand within the healthcare system. In
addition, the lack of radiologists makes it difficult and time-
consuming to analyse medical pictures. Using artificial
intelligence (AI), these issues can be solved. Machine
Learning (ML) is an application of AI that can learn without Fig 1 Technologies used
having to be specifically programmed, that learns from data,
and that makes predictions or judgements based on  X-Ray Image:
historical data.ML makes use of supervised learning, Chest radiography is mainly used in diagnostic
unsupervised learning, and semi-supervised learning, three procedures to detect conjecture heart failures and lung
types of learning advancements. The ML approaches diseases such as tuberculosis, atelectasis, asthma, pleural
involve feature extraction and the choice of feature effusion, pneumothorax, hyper cardiac inflation, and
selection. pneumonia. X-ray images are accessible, inexpensive, and
less dose-effective compared to other imaging processes,
and it is a powerful tool for mass examination. Deep

IJISRT23MAY270 www.ijisrt.com 213

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
learning methods used for X-ray image analysis. S. Hwang  Mammograph (MG):
et al. implemented the first deep CNN-based Tuberculosis One of the top causes of death for cancer-stricken
screening system with a transfer learning technique. women worldwide is breast cancer. The most popular and
Rajaraman et al. proposed modality-specific ensemble reliable method for early detection of breast cancer is
learning for the detection of anomalies in chest X-rays magnetic resonance imaging (MG). A low-dose x-ray
(CXRs). These model predictions are combined using many imaging technique called MG is utilised to see the breast's
ensemble techniques toward minimizing prediction structure in an order to find breast illnesses.The tumors
variance. Class selective mapping of interest (CRM) is used make up a little portion of the actual breast picture, making
for visualizing the normal regions in the CXR images. it challenging to identify breast cancer on mammography
screenings. There are three processes in the analysis of
breast lesions from MG: detection, segmentation, and
classification. Still major topic of research is the automatic
classification and early mass detection in MG. DLA has
demonstrated some major breast cancer progress during the
previous ten years.

Fig 2X-ray Image

 Computerized Tomography (CT):

A special computer program processed this large
volume of data to create two-dimensional cross-sectional
images of our body.This imaging test helps to detect internal
injuries and diseases by providing cross-sectional images of Fig 4 Mammograph
bones, blood vessels,and soft tissues. CT is a high detection
capability, reveals small lesions, and provides a more  Histopathology:
detailed assessment.CT scanning is frequently used for lung Histopathology is the examination of human tissue
nodule identification. The detection of malignant pulmonary under a microscope and a sliding glass to diagnose
nodules is fundamental to the early diagnosis of lung cancer conditions such as kidney, lung, breast, and other cancers.
Table 4 summarizes the latest deep-learning developments Staining is used in histopathology to visualise and highlight
in the study of CT image analysis. Lietal.2016 implemented a specific region of the tissue. For instance, the nucleus is
to solve the recognition of three types of nodules, that is, stained with hematoxylin and eosin (H&E), which gives
solid, semi-solid, and ground glass Balagourouchetty et al. other structures a pink tone and the nucleus a dark purple
introduced Google Net based on ensemble FC Net classifier hue. A H&E stain has been essential in the past century for
for liver lesion classification.Masood et al. implemented the identifying various illnesses, diagnosing cancer, and
multidimensional Region-based Fully Convolutional grading. Modern imaging technology includes digital
Network(RFC) for lung nodule detection/classification and pathology. Deep learning is shown promise in the analysis
achieved classification accuracy of 97.91%. In lung nodule of histopathology pictures, particularly in the areas of
detection, the feature work is the detection of micronodules nucleus recognition, image categorization, cell
(less than 3 mm) without loss of insensitive and accurate. segmentation, tissue segmentation, etc.

Fig 5 Histopathology
Fig 3 Computerized Tomography

IJISRT23MAY270 www.ijisrt.com 214

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
III. SOFTWARE REQUIREMENTS  Performance:
SPECIFICATION Performance is assessed using following
specifications:
A. Functional Requirements:
The requirements that specify what all services system
 Response time:
can provide the end-user is called the functional
requirements. These define exactly what functions the It is the time taken for the system to accept user input
system can do. The functional requirements are closely and respond to it by displaying some output. The response
related to the user requirement specifications. This may time must also be consistent and not vary based on the
include calculations, data processing, technical operations number of concurrent sessions.
and other such functionality that aim to fulfill the
application objectives.  Workload:
It refers to how much stress or work the system can
These are captured in the form of use cases, which are
handle simultaneously. This could be in terms of parallel
the system responses to events by external agents or internal
deadlines. Any tracking operations, legal requirements, sessions, number of active users or number of database
interface details, authorization levels, transaction, updates, transactions.
and cancellations, and administrative functions come under
functional requirements. The technical architecture of the  Throughput:
system is determined by these requirements. The number of samples or bytes of data that are
processed per second is referred to as throughput. The data
B. Non Functional Requirements: processing rate should be as high as possible to ensure that
the outputs are consistent and the user sustains interest in
using the system.

 Reliability: Based on the outcomes of :

 Integrity: The information has not been modified by

non-authorised people.

 Authenticity: A proof that the information belongs to

the correct patient and issued from the right authority.

 Availability: Warrants an information system to be used

in the normal schedule conditions of access.

IV. EXISTING SYSTEM

Most imaging procedures need the patient to remain

still while photos are being taken. Since a child only needs
Fig 6 Non Functional Requirements to be still for a brief period of time for an X-ray or an
ultrasound procedure, we frequently avoid sedated or
 Extensibility: restrained older children during X-ray procedures. However,
The design principle that determines the ability of a smaller children and those who are anxious in unfamiliar
surroundings may require assistance lying motionless,
system to be extended is called extensibility. The expansion typically from parent.
could consist of new functionality or a change to already-
existing functionality. Overall, the system is enhanced while  Limitations of Existing System
not affecting existing working functions.
 Medical imaging complications are uncommon, but
 Performance: they can be serious, possibly resulting in an injury or
It is how a system works or performs by taking the causing a secondary illness.
input training data and testing it and then classifying and  Most imaging procedures require the patient to lie still
for longer periods of time.
predicting.
 The diagnosis of a disease is delayed because imaging
could not be performed.
 Interface:  The diagnosis of image took more time so that the
This application interacts with the doctor and show the person becomes sick.
diagnosed medical images.

IJISRT23MAY270 www.ijisrt.com 215

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
V. PROPOSED SYSTEM  Random Forest:
Popular machine learning algorithm Random Forest is
In the proposed method, we have enhanced medical a part of the supervised learning methodology. It can be
images by using effective enhancement algorithms, which used for ML problems involving both regression and
are the median filter, unsharp mask filter, and contrast- classification. It is based on the concept of ensemble
limited adaptive histogram equalization. The proposed learning, which is a technique for combining several
methods have been implemented by MATLAB and more classifiers to handle challenging problems and improve
than 60 medical images in Different parts of the body have model performance.
been used to evaluate the performance of the proposed
methods. Depending on the medical imaging modalities, As its name suggests, Random Forest is a classifier
input medical images can be improved by up to 80%. This that averages several decision trees applied to various
outcome is based on the assessments of professionals who subsets of the supplied information to improve the predicted
compared input and output photographs. accuracy of the dataset. The graph below shows how the
Random Forest algorithm works.
A. Advantages of Proposed System:

 Algorithms have been proposed in this approach.

 The patient need does not wait for longer time.
 The result of the diagnosed image is based on the
training data.
 The accuracy is more because the system works more
accurately.

B. Machine Learning:
With the use of machine learning (ML), which is a
form of artificial intelligence (AI), software programmers
can predict outcomes more accurately without having to be
explicitly instructed to do so. In order to forecast new output
values, machine learning algorithms use historical data as
input. Fig 8 Random Forest

C. Algorithm Used:  K-Nearest Neighbour:

One of the most fundamental supervised learning-
 Support Vector Machine: based machine learning algorithms is K-Nearest
To deal with classification and regression issues, the Neighbor.The K-NN algorithm places the new instance in
Support Vector Machine (SVM),is one of the most well- the category that resembles the current categories the most,
liked supervised learning techniques, is used. However, the presuming that the new case and the previous cases are
majority of its makes use of are in Machine Learning comparable. After storing all the previous data, new data
Classification problems. point is categorised using the K-NN algorithm based on
similarity. This indicates that new data can be reliably and
In an order to create the hyperplane, SVM selects the quickly categorised using the K-NN approach.Although the
extreme points and vectors. Support vectors, which are used K-NN approach is most frequently employed for
to represent these extreme instances, are what to give the classification problems, it can also be utilised for regression.
Support Vector Machine method its name. Take a look at
the diagram below, where decision boundary or hyperplane
is used to categorize two distinct categories.

Fig 7 Support Vector Machine Fig 9 K-Nearest Neighbour

IJISRT23MAY270 www.ijisrt.com 216

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
 Convolutional Neural Network: that aim, researchers in the field of medical imaging are
A class of deep neural networks called convolutional looking for new methods to use technology and computing
neural networks (CNN/ConvNet) are most frequently used power to break down existing boundaries. Real-time, less
to assess visual imagery. When we think of neural networks, invasive methods of observing processes like tumour growth
matrix multiplications typically spring to a mind, yet and cell division are sought for. Like medical imaging has
ConvNet is not a neural network. It employs a special traditionally done, further future advancements will adapt
technique known as convolution. Convolution is a present technology and practises to provide enhancements.
mathematical procedure that takes two functions and creates
a third function that expresses how the shape of one is VI. CONCLUSION
changed by the other in mathematics.
Without evasive surgery, medical imaging would not
be able to reveal anything about the human body or the
problems that surround it.Diseases may be easier to treat
than ever before thanks to medical imaging. Future
advancements in medical imaging technology appear to be
inevitable..

Diagnostic imaging has mostly been utilised to assess

individuals with unusual clinical presentations.
Improvements in imaging technology over the past few
years have helped these patients' diagnoses more accurately.

Fig 10 Convolutional Neural Network REFERENCES

 System Architecture: [1]. Abadi M et al. (2016) TensorFlow: Large-Scale

Machine Learning on Heterogeneous Distributed
Systems,
[Online].Available:http://arxiv.org/abs/1603.04467.
[2]. Abbas A, Abdelsamea MM, Gaber MM (2020)
Classification of COVID-19 in chest X-ray images
using DeTraC deep convolutional neural network, pp.
1–9, [Online]. Available: http://arxiv.org/abs/ 2003.
13815.
[3]. Apostolopoulos ID, Mpesiana TA (2020) Covid-19:
automatic detection from X-ray images utilizing
transfer learning with convolutional neural networks,
Phys Eng Sci Med, no. 0123456789, pp. 1–6, DOI:
https://doi.org/10.1007/s13246-020-00865-4.
[4]. Arun NT et al. (2020) Assessing the validity of
saliency maps for abnormality localization in medical
imaging, pp. 1–5, [Online]. Available:
http://arxiv.org/abs/2006.00063.
[5]. L.Balagourouchetty, J.K. Pragatheeswaran, B.
Pottakkat, and R. G, “GoogLeNet based ensemble
FCNet classifier for focal liver lesion diagnosis,”
IEEE J Biomed Heal Inf, vol. 2194, no. c, pp. 1–1,
2019, DOI:https://doi.org/10.1109/jbhi. 2019.
2942774, 1694.
[6]. Bastien F et al. (2012) Theano: new features and
speed improvements, pp. 1–10,[Online]. Available:
http://arxiv.org/abs/1211.5590.
[7]. Basu S, Mitra S, Saha N (2020) Deep Learning for
Screening COVID-19 using Chest X-Ray Images, pp.
1–6, [Online]. Available: http://arxiv.org/abs/ 2004.
10507
Fig 11 System architecture [8]. Bauer S, Wiest R, Nolte LP, Reyes M (2013) A
survey of MRI-based medical image analysis for
 Future Scope: brain tumor studies. Phys Med Biol 58(13):1–44.
Medicine has been transformed by imaging, and this https://doi.org/10.1088/00319155/58/13/R97
trend is expected to continue quickly as more information
about each human biological system becomes available. To

IJISRT23MAY270 www.ijisrt.com 217

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
BIOGRAPHIES

MR.BORUGUDDA. NAGA RAJU is currently

working as a Associate Professor in the department of
Information Technology from NRI Institute of
Technology,Pothavarappadu, Andhra Pradesh. He received
M.Tech from JNTU KAKINADA and he published various
national and international Journals. Heis a member of CSI.
He is a ratified faculty from JNTU KAKINADA.

R.Srija is currently studying B.Tech with specification

of Information Technology in NRI Institute of Technology.
She done a summer internship Diagnoising medical images
and she done 2 NPTEL certifications.

M.Komala is currently studying B.Tech with

specification of Information Technology in NRI Institute of
Technology. She done a summer internship Diagnoising
medical images and she done 2 NPTEL certifications.

P.Dharani is currently studying B.Tech with

specifiaction of Information Technology in NRI Institute of
Technology.She done a summer internship Diagnoising
medical images and she done 2 NPTEL certifications.

IJISRT23MAY270 www.ijisrt.com 218