Convolutional Neural Network based Lungs

diseases prediction
Prof. Seema Kedar Ms. Sanika Shirsat
Computer department Computer department
JSPM’s Rajarashi Shahu College of Engineering JSPM’s Rajarashi Shahu College of Engineering
Pune, India Pune, India
[email protected] [email protected]

Abstract---Disease Recognition Systems have played a air sacs filled with pus or fluid maybe in one or both
vital role in history all over world and is widely found in lungs [2]. The most dangerous diseases and the
today’s time. Various researchers have worked on higher rate of death is occurred by lung cancer. This
different diseases using different algorithms such as
disease has uncontrollable cell growth in lung tissues.
skin diseases, breast diseases, blood issues, cancers,
88% of lung cancer is caused due to long term
tumors, etc. Research efforts on lungs related diseases
such as lung cancer, pneumonia and covid-19. This smoking. And rest is due to asthma, asbestos, or
research intends to bring a system which can recognize radon gas. To check the presence of lung cancer CT-
lungs diseases. Methods based on CNN classifier are scan or X-rays can be helpful [3]. In this work we
exercised for Lung Diseases Recognition. Experiments show how local phase chest x-ray features based
are carried out on datasets using different kinds of image enhancement improves the accuracy of CNN
features and their fusion. architectures for COVID19, pneumonia and lung
cancer diagnosis. Also, CT- scan for lung cancer.
Keywords---Lung Cancer, Covid-19, Pneumonia, Normal,
Convolutional Neural Network

I. INTRODUCTION
The research is based on Lungs Disease Recognition.
As we know, there are many diseases related to
lungs. Using prediction error results are reported,
which is inverted classification Accuracy. The
coronavirus disease 2019 (COVID-19) caused by
severe acute respiratory syndrome coronavirus 2
Fig. 1. X-Ray Images of lungs
(SARS-CoV-2) is one of the highly destructive
infectious diseases after ages. This new type of Specifically, we exact different features of Chest X-
diseases was announced in late December, 2019 and ray combined with multi-feature image. CNN
World Health Organization (WHO) declared it as a architecture will be used for processing multi-feature
pandemic, affecting countries worldwide. Numerous chest x-ray data. Our proposed methods will evaluate
studies have shown great results in identifying Covid- disease on large scale based on X-ray images
19 using X-ray images. Furthermore, most of the obtained from healthy subjects as well as subjects
previous work was evaluated on a limited number of who are diagnosed with community acquired lung
COVID19 chest x-ray scans [1]. According to the cancer, pneumonia and COVID-19. Quantitative
report of 2015, 920.000 toddlers were killed due to results show the usefulness of local phase image
pneumonia. The symptoms identified are fever, features for improved diagnosis of COVID-19
chills, breathing difficulty, and cough with pus also disease from CXR scans.
 II. LITERATURE SURVEY TABLE I. LITERATURE SURVEY
This section contains the review of the previous work
Work Reference Year Technique Result
carried out. We will get to know the existing work (%)
taken place on the different types of images for
Anand Nayyar et al. [4] 2020 Resnet 50 94%
diagnosing lungs related disease.
Matias Cam Arellano et 2020 CNN 94.79%
al. [5]
Nayyar et al. [4], states that swelling of tissues in one
or both lungs and the diseases which can be viral, Xi Ouyang et al. [6] 2020 Dual Sampling 87.5%
Strategy
bacterial or fungal is Pneumonia. Due to human
errors the diseases can’t be predicted appropriately Grega Vrbancic et al. 2020 CNN with 94%
which leads to death of the patient. Using Mask [7] optimized
tuning
RCNN object detector infected regions can be
Harsh Sharma et al. [8] 2020 CNN 90%
identified and diagnosed using this system. 0.155 is
the IoU score obtained. Matias Cam Arellano et al. Nidhi S. Nadkarni et al. 2019 SVM 92%
[9]
[5], the diseases having the current highest rate of
deaths in Covid-19. CNN is the algorithm used by the Kvamelia Roy et al. [10] 2019 SVM 94.5%

system to detect the disease. Xi Ouyang et al. [6], Zhong liang Li et al. 2019 PNet 92.79%
focuses on Covid-19 from pneumonia using Dual [11]
Sneha Potghan et al. 2018 KNN, MLP 98.30%,
Sampling. From approximately 8 hospitals the data is [12] 98.31%
being collected of different age groups. In training
stage 5-fold cross validation is applied. Grega
Vrbancic et al. [7], dataset holding 842 X-ray images
of chest. The algorithm used is grey wolf optimizer III. EXISTING SYSTEM
with CNN for optimizing hyper-parameter values. The main objective of the project was to find an
Accuracy obtained is 94.76%. In [8] Harsh Sharma et efficient algorithm for feature extraction with
al. states, the traditional methods are pretty time reduced length of feature vector as well as which can
consuming. To identify the patient is attacked by increase the accuracy of recognition of X-ray images.
pneumonia a system is designed using CNN For doctors it is very easy to recognize or read x-rays
algorithm. Dataset holds original as well as as human beings do not recognize things on the basis
augmented images. Nidhi S. Nadkarni et al. [9] of extract structure of pattern of any entity. We have
proposed, cancer detection using SVM algorithm on the ability to derive the common patterns very
the CT-scan images. Here for preprocessing median quickly through the intelligence gifted by God. But to
filtering is used while for segmentation mathematical write such an intelligent program for computer to
morphological operation is used. In [10] Kvamelia show such characteristics like humans is a difficult
Roy et al., Covid detection is done using SVM and a tedious job because computer works on
combined with saliency enhancement. DICOM comparison if the two entities are exactly the same it
format dataset is being used for the same. Zhong can identify otherwise it will not. To make the
liang Li et al. [11], for detecting Pneumonia 10784 computer intelligence it is necessary to first analyze
images of chest X-ray dataset is used with PNet deep and detect the procedure following which human gets
learning-based framework. Sneha Potghan et al. [12], able to remind or recall or identify things i.e., the
used KNN with Multi-layer perceptron algorithm for uniqueness among them which is referred as feature
prediction Covid-19. vector. Then the next task is to make the computer
learn it to be able to make judgements by adjusting
the values.
 IV. SYSTEM ARCHITECTURE REFERENCES

[1] Rachna Sethi, Monica Mehrotra, Dhaarna Sethi, “Deep

Learning based Diagnosis Recommendation for COVID-19
using Chest X-Rays Images”, 2020 IEEE
[2] Abdullah Faqih Al Mubarok, Dominique Jeffrey A. M.,
Ahmad Habbie Thias, “Pneumonia Detection with Deep
Convolutional Architecture”, 2019 IEEE
[3] Heng Yu, Zhiqing Zhou, Qiming Wang, “Deep Learning
Assisted Predict of Lung Cancer on Computed Tomography
Images Using the Adaptive Hierarchical Heuristic
Mathematical Model”, 2020 IEEE
[4] Anand Nayyar, Rachna Jain, Yash Upadbyay, “Object
detection-based approach for Automatic detection of
Pneumonia”, 978- 1·728 1-6770- 1/20/$31 .00 0 2020 IEEE
Fig. 2. System Architecture
[5] Mat´ıas Cam Arellano, Oscar E. Ramos, “Deep Learning
Model to Identify COVID-19 Cases from Chest
We have proposed an end-to-end trainable deep Radiographs”, 2020 IEEE
CNN. Using Chest X-ray images great extent of work [6] Xi Ouyang† , Jiayu Huo† , Liming Xia† , Fei Shan† , Jun
has already been done for recognition diseases with Liu† , Zhanhao Mo† , Fuhua Yan† , Zhongxiang Ding† , Qi
different CNN algorithms. This is one of the biggest Yang† , Bin Song† , Feng Shi, Huan Yuan, Ying Wei,
Xiaohuan Cao, Yaozong Gao, Dijia Wu, Qian Wang∗ ,
challenges with traditional CNN algorithm because a Dinggang Shen∗, “Dual-Sampling Attention Network for
programmer needs to tell the computer in advance Diagnosis of COVID-19 from Community Acquired
what features should be taken care of so that to get an Pneumonia”, 2020 IEEE

accurate decision percentage to recognize the [7] Grega Vrbancic, Spela Pecnik, Vili Podgorelec,
“Identification of COVID-19 X-ray Images using CNN with
diseases based on the X-rays and CT-Scan. Normally Optimized Tuning of Transfer Learning”, 978-1-7281-6799-
Disease recognition using CNN is divided into six 21201$31.00 ©2020 IEEE
phases which are acquisition, pre-processing, [8] Harsh Sharma, Jai Sethia Jain, Priti Bansal, Sumit Gupta,
“Feature Extraction and Classification of Chest X-Ray
segmentation, feature extraction, classification and Images Using CNN to Detect Pneumonia”, 978-1-7281-
post processing. Feature Extraction plays an 2791-0/20/$31.00 c 2020 IEEE
important role in getting the result of a model. The [9] Nidhi S. Nadkarni, Prof. Sangam Borkar, “Detection of Lung
purpose behind CNN is to identify and analyze a Cancer in CT Images using Image Processing”, 978-1-5386-
9439-8/19/$31.00 ©2019 IEEE
document image by dividing it into disease. These [10] Kyamelia Roy, Sheli Sinha Chaudhury, Madhurima Burman,
diseases are further compared with image pattern to Ahana Ganguly, Chandrima Dutta, Sayani Banik, Rayna
predict the probable disease. With the advent of Banik, “A Comparative study of Lung Cancer detection using
supervised neural network”, 978-1-7281-0070-8/19/$31.00
highest computing power and deep learning,
©2019 IEEE
applications of computer vision like disease [11] Zhongliang Li12† , Juan Yu45† , Xuechen Li12, Yingqi Li6 ,
recognition took a little relief because deep learning Weicai Dai7 , Linlin Shen12*, Lisha Mou34* and Zuhui Pu4,
models are itself responsible for what features to look “PNet: An Efficient Network for Pneumonia Detection”,
978-1-7281-4852-6/19/$31.00 ©2019 IEEE
upon without explicitly guided Convolutional Neural
[12] Sneha Potghan, R. Rajamenakshi, Dr. Archana Bhise,
Network. “Multi-Layer Perceptron based Lung tumor classification”,
978-1-5386-0965-1/18/$31.00 ©2018 IEEE