1. Supervisor
Dr. S. M. ALI
Principal
ACET, Nagpur
ANJUMAN COLLEGE OF ENGINEERING &
TECHNOLOGY, NAGPUR
Classification of Images for Diagnosis of Lung Infection using CNN
In
M. TECH. ( ELECTRONICS & COMMUNICATION ENGINEERING )
Presented By -
Mr. Harsharatna. M. Kosey
2. Outlines
▪ Introduction
▪ Literature Review
▪ Problem Defination
▪ Research Objective
▪ Research Methodology (Flow Chart / Block
Diagram)
▪ Application
▪ Propose Plan of Work
M. TECH. ( ELECTRONICS & COMMUNICATION ENGINEERING )
2
3. Introduction
⚫ We propose a method to treat different images that share common properties found in
patient metadata as positive pairs in the context of contrastive learning. We
demonstrate the application of this method to a chest X-ray interpretation task
⚫. Similar to the concurrent work by we experiment with requiring positive pairs to
come from the same patient as these images likely share highly similar visual features.
However, our method incorporates these positive pairs with possibly different images
directly as part of the view generation scheme in a single contrastive pre-training stage,
as opposed to which adds a second pre-training stage where a positive pair must be
formed by two distinct images
3
M. TECH. ( ELECTRONICS & COMMUNICATION ENGINEERING )
4. 4
Sr no Paper Title &
Authors
Details of publications Finding
01 ]Singh and Gupta [2] Performance
analysis of various machine
learning-based approaches for
detection and classification of
lung cancer in humans
Neural Comput Appl
2019;31(10):6863–77
applied Relu based deep
learning method in identifying
the malignant lung cancer from
the image data set, their
detection rate is 85.55%.
02 Liu J, Pan Y, Li M, Chen Z, Tang L,
Lu C, etal[3] Deep learning and
medical imaging. J Thorac
Dis 2018;10(3):1325. For this, deep
learning/machine learning/data
mining classifiers have been
immensely applied in order to
extract the relevant features
from image data sets and
classify them for disease
diagnosis and prediction
03 Rajpurkar et al. [4]. . Lung
infection quantification of covid-
19 in ct images with deep
learning.
arXiv preprint
arXiv:2003046552020.
Similarly, in the current
scenario of COVID-19, a CNN
based approach has been
applied to the X-ray images of
the chest, since it causes
similar symptoms to
Pneumonia, verily, more
severe.
An automated COVID-19
method has been proposed by
Literature Review/Survey
5. 5
Sr no Paper Title &
Authors
Details of publications Finding
04 Rajpurkar et al. [4].
Applications of deep learning
to mri images: asurvey.
Big Data Min Anal
2018;1(1):1–18.
Similarly, in the current
scenario of COVID-19, a
CNN based approach has
been applied to the X-ray
images of the chest, since it
causes similar symptoms to
Pneumonia, verily, more
severe.
05 An automated COVID-19 method
has been proposed by Alqudah et
al. [5]
Chexnet: radiologist-level
pneumonia detection on chest
x-rays with deep learning.
arXiv preprint
arXiv:1711052252017.
they used machine learning
classifiers–Support Vector
Machine (SVM), random
forest, K-Nearest Neighbour
(KNN) with the CNN using
soft-max, this helps in
detection with a high
accuracy rate of 98%.
06 Esteva et al. [6Automated
systems for detection of
covid-19 using chest x-ray
images and lightweight
convolutional neural networks
2020; 10.21203/rs.3.rs-
26500/v1.
] focused on manifesting the
classification of skin lesions,
a single CNN layer is used,
for diagnosis from diverse
clinical images of skin. The
model was tested to classify
them binary with two critical
cases– common cancer and
deadly cancer. Further, its
application also seen in
tuberculosis data sets,
6. Problem Statement/Hypothesis
⚫objective is to examine if transfer learning
improves a model’s performance when using a
small dataset consisting of radiographs. In this
case, the model’s goal is to classify healthy,
pneumonia and COVID19.
⚫Can transfer learning from different source
domains improve performance for classifying
COVID-19 - Does it matter if the basis for the
transfer learning is within the same domain or no
6 M. TECH. ( ELECTRONICS & COMMUNICATION ENGINEERING )
7. Research Objectives
•The goal of the project is to find out if Deep transfer learning using pneumonia
dataset is a valid approach to build a model that classifies patients successfully.
This is a highly relevant issue considering the recent outbreak of COVID-19, as
stated previously.
Furthermore, it can also be useful to find approaches for creating machine
learning models with limited training data, for other medical diagnostics
problems. As in the medical domain, labeling of data often requires knowledge
from skilled professionals which can be tedious.
⚫ The aim of this study is to develop and test a reliable diagnostic tool, using
deep learning technology to detect COVID-19 features from chest X-rays. This
tool would accelerate the diagnosis and referral of patients, improving clinical
outcomes.
⚫ Releasing the deep learning model as open source would facilitate the use of
the tool both now and in any future pandemics, where a similar algorithm
could be used.
⚫ This particular application also has the potential to be scaled up and used for
more generalized high-impact applications in biomedical imaging
7 M. TECH. ( ELECTRONICS & COMMUNICATION ENGINEERING )
9. 9
Applications
application of deep learning in the field of COVID-19 radiologic image processing reduces false-
positive and negative errors in the detection and diagnosis of this disease and offers a unique
opportunity to provide fast, cheap, and safe diagnostic services to patients.
used a dimensionality reduction method to generate a set of optimal features of CXR images
to build an efficient machine learning classifier that can distinguish COVID-19 cases from non-
COVID-19 cases with high accuracy and sensitivity.
10. Propose Plan of Work
10
OCT-NOV
(2021)
1)Literature Review
2)Synopsis Submission
3)Detail and study of Project
DEC-JAN
(2021-2022)
Implementation of Project
And Review Paper
FEB-MAR
(2022)
Analysis of result and Error
Correction
FEB-MAR
(2022)
Final Paper / Report/Thesis
Writing
11. References
⚫ [1] Diederik P. Kingma*, Jimmy Lei Ba*. “ADAM: A METHOD FOR STOCHASTIC
OPTIMIZATION”. In: (2015). URL: https://arxiv.org/pdf/1412.6980.pdf.
⚫ [2] Facial Expression Recognition Challenge. 2013. URL: https:// www. kaggle. com/
c/challenges- in- representation- learning- facial- expression- recognition-
challenge/overview.
⚫ [3] Goossens, Michel, Mittelbach, Frank, and Samarin, Alexander. ”The Companion”.
Reading, Massachusetts: Addison-Wesley, 1993.
⚫ [4] Hong-Wei Ng, et al. “Deep Learning for Emotion Recognition on Small Datasets Using
Transfer Learning”. In: (2015). URL: http://dx.doi.org/10.1145/2818346.2830593.
⚫ [5] Jason Wang, Luis Perez. “The Effectiveness of Data Augmentation in Image
Classification using Deep Learning”. In: (2017). URL: https://arxiv.org/pdf/1712. 04621.pdf.
⚫ [6] Joseph Paul Cohen Paul Morrison, Lan Dao. “COVID-19 Image Data Collection”. In:
(2020). URL: https://arxiv.org/pdf/2003.11597.pdf.
⚫ [7] Karen Simonyan, Andrew Zisserman. “VERY DEEP CONVOLUTIONAL NETWORKS
FOR LARGE-SCALE IMAGE RECOGNITION”. In: (2015). URL: https://arxiv.org/
pdf/1409.1556.pdf.
⚫ [8] Mooney, Paul. Chest X-Ray Images (Pneumonia). 2018. URL: https://www.kaggle.
com/paultimothymooney/chest-xray-pneumonia.
⚫ [9] Murphy, Kevin P. Machine Learning: A Probabilistic Perspective (Adaptive
Computation and Machine Learning series). ”The MIT Press; 1 edition ”, 2012.
⚫ [10] Organization, World Health. In: (2020). URL: https://www.who.int/publications-
detail/covid- 19- operational- guidance- for- maintaining- essential- health- services-during-
an-outbreak.
11 M. TECH. ( ELECTRONICS & COMMUNICATION ENGINEERING )
12. Methodology
⚫ We build a novel architecture to classify the input COVID-19 CXR images into normal and
abnormal categories. The proposed network is called Chest X-Ray COVID-19 Network
(CXRVN) is considered as the first specialized deep neural network for analyzing chest X-ray
images against the pandemic COVID-19. Our network architecture is summarized in Fig. 2.
Generally, CXRVN consists of four convolution layers, three pooling layers, and one fully
connected layer. Next, we describe the main features of our architecture and their
importance for diagnosis COVID19 patients.
⚫ The collected datasets consist of normal cases and COVID-19 ones. These datasets are
splitted into two sets, and they are the training and testing sets. To overcome the overfitting
problem, We split the datasets into 80% for the trained images, and the remaining 20% is for
the testing ones. Subsequently, the training sets are augmented via the use of GAN.
Therefore, the hyper-parameters values of the training sets have learned and proceeded with
the evaluation to produce the validation set. Every iteration of the shuffled fold is split by
generating an independent number of the trained/tested image.
12 M. TECH. ( ELECTRONICS & COMMUNICATION ENGINEERING )
14. Result
⚫ In this study, we had trained a network from scratch on CXR images and classified them
into COVID+VE and COVIDVE classes. We had tested on CXR images that were not used
in the training phase, and got good results that were 95:63% accurate(overall). We had
trained the model upto 1000 iterations which we found sufficient to validate the
trustworthy of our results.
14 M. TECH. ( ELECTRONICS & COMMUNICATION ENGINEERING )
Training and validation accuracy of the mode
15. Performance Study
⚫ Let's measure the performance of our algorithm in terms of confusion matrix - This
metric also gives a good idea of the performance in terms of precision and recall. We
believe overall accuracy is a good indicator as the testing dataset utilized in this study is
uniformly distributed
15 M. TECH. ( ELECTRONICS & COMMUNICATION ENGINEERING )
Confusion Matrix. Green colored region depicts correctly classified data and orange colored
region depicts wrongly classified data
16. Future Enhancement
16 M. TECH. ( ELECTRONICS & COMMUNICATION ENGINEERING )
⚫ ImageNet and radiographs were tested for transfer learning
and the results were quite interesting. It may turn out that the
best performing model uses transferred weights from a
completely different domain.
⚫ It would therefore be interesting to examine different domains
and apply different network structures such as DenseNet or
ResNet. In network N1, which formed the source network for
transfer to N3, it would be interesting to use the same
architecture
17. Application
17 M. TECH. ( ELECTRONICS & COMMUNICATION ENGINEERING )
⚫ Hospitals are limited on resources with an overflow of patients,
making the current amount of healthcare professionals
insufficient to test and attend to all patients
⚫ A potential solution for testing patients quicker could be a
machine learning based model, similar to those currently in
use for other diseases like breast cancer and pneumonia