This document discusses detecting pneumonia in chest X-rays using deep learning techniques. It begins by stating that pneumonia is a major cause of death worldwide, especially in children. The objective is to develop a deep learning framework to automatically diagnose pneumonia from chest X-rays to reduce human error. Various deep learning models like CNN, VGG-16 and MobileNetV2 are implemented and compared on a public dataset. VGG-16 achieved the highest accuracy of 94.3% among the models for detecting pneumonia. The document concludes that pneumonia can be identified and classified using deep learning models with VGG-16 performing best.

1. DETECTING PNEUMONIA IN
CHESTX-RAYSUSING DEEP
LEARNINGTECHNIQUE
Project Guide:
Dr.T. Jemima Jebaseeli
Done by:
Punithavathy S P
PRK19CS1011

2.  India accounts for 20 percent of the death worldwide caused by
pneumonia
 Pneumonia affects the lungs and causes about 18% of all deaths in
children under five years old.
 Chest X-rays are the most common and easiest way to detect
pneumonia.
 X-ray images are often unclear and can be confused with other
diseases that have similar features.
Problem
Statement

3.  To reduce the human eye error in diagnosing the disease.
 To develop a deep learning framework to automatically diagnose
pneumonia using chest X-ray images.
 To classify the result as normal cases or pneumonia cases.
Objective

4.  Machine Learning library ,sklearn used with many subordinate
models such as model classifiers, and metrics etc.
 Deep Learning library: Keras, Pytorch etc..
 Google Colab used for implementation.
 Data set available on open source platforms
 Programming Language: Python
TECHNIQUES/
TECHNOLOGIES
USED

5.  The dataset is taken from kaggle.
 The dataset is organized into 2 folders (train, test) and contains
subfolders for each image category (Pneumonia/Normal).
Dataset

6. Methodology:
IMAGEACQUISITION
IMAGE PREPROCESSING
SEGMENTATION
FEATURE EXTRACTION
NORMAL CASES PNEUMONIA CASES
CNN
CLASSIFICATION

7. Preprocessing
Preprocessing
 Preprocessing is the removal of systematic noise from the data
 Image generator object which performs random rotations,
shifts, flips, crops, and sheers on the image dataset.
Enhancement
• Image enhancement is the process of adjusting digital images so
that the results are more suitable for display or further image
analysis.
• Enhancement remove noise, sharpen, or brighten an image,
making it easier to identify key features.

8. Preprocessing
NOISE REMOVAL:
• Digital images are prone to various types of noise.
Noise is the result of errors in the image acquisition process that
result in pixel values that do not reflect the true intensities of the real
scene
There are several ways that noise can be introduced into an image,
depending on how the image is created. Electronic transmission of
image data can introduce noise.

9. SEGMENTATION:
• Image segmentation is the process of dividing an image
into multiple parts.
• Simplify and easier to analyze.
• This is typically used to identify objects or other
relevant information in digital images.
segmentation:

10.  CNN has two parts Feature extraction and classification
 A Convolutional Neural Network (CNN) is a deep learning
algorithm that can recognize and classify features in images for
computer vision.
 A convolution tool that splits the various features of the image for
analysis
 Feature extraction - several convolution layers followed by max-
pooling and an activation function.
 The classifier -fully connected layers.
Overview of -
CNN

11. CNNArchitecture

12.  Convolutional layer works by placing a filter over an array of image
pixels.
 Convolved feature map
Convolution
layer

13.  Redues the sample size
 Makes the process much faster
 Pooled feature map
Two types
 Max pooling
 Average pooling
Pooling Layer

14. CONVOLUTIO
NAL NEURAL
NETWORK
Layers:
 Convolution Layer
 ReLu Layer
 Pooling Layer
 Fully Connected Layer

15. Traning and
testing

16. Accuracy  Accuracy , precision and recall on testing set:

17. VGG-16

18. VGG-16
 VGG stands for Visual Geometry Group
 TheVGG architecture consists of blocks, where each block is
composed of 2D Convolution and Max Pooling layers.
 There are total of 13 convolutional layers and 3 fully connected
layers inVGG16 architecture.
 VGG-16 is a convolutional neural network that is 16 layers deep.
 The network has an image input size of 224-by-224.

19. Architecture

20. VGG-16 layers

21. Testing and
traning

22. Accuracy

23. Mobile NetV2

24. Mobile net
 MobileNetV2 is a convolutional neural network architecture that
seeks to perform well on mobile devices.
 It is based on an inverted residual structure where the residual
connections are between the bottleneck layers
 The architecture of MobileNetV2 contains the initial fully
convolution layer with 32 filters, followed by 19 residual
bottleneck layers.
 ...

25. Architecture

26. Mobile net
layers

27. Testing and
traning

28. Accuracy

29. Output
Comparision
CNN Model
Accuracy = 90.6%
VGG-16
Accuracy = 94.3%
Mobile net model
Accuracy = 85.00%

30. INFERENCE
 •CNN model for Pneumonia disease detection –implemented.
 •Mobile Net model Pneumonia Pneumonia Disease detection –
implemented.
 VGG-16 model for Pneumonia disease detection –implemented
 Classification of all models –tested.
 VGG-16 outperformed Mobile Net model, CNN Model for
pneumonia disease detection in Chest X-rays.

31. CONCLUSION
 •Pneumonia identified using Deep learning models.
 •Pneumonia Disease was classified using Deep learning models.
 •CNN model ,VGG-16 and Mobile net was used and compared for
accuracy and performance.
 •VGG-16 model proved to produce accuracy than CNN Model and
Mobile net for detecting Pneumonia disease in Chest X-rays.

32.  [1] U. Avni, H. Greenspan, E. Konen, M. Sharon, J. Goldberger, X-ray
categorization and retrieval on the organ and pathology level, using patch-
based visual words, IEEE Trans. Med. Imaging 30 (3) (2011) 733–746
.
 [2] P. Pattrapisetwong, W. Chiracharit, Automatic lung segmentation in
chest radiographs using shadow filter and multilevel thresholding, in: 2016
International Computer Science and Engineering Conference (ICSEC),
IEEE, 2016, pp. 1–6.
 [3] S. Katsuragawa, K. Doi, Computer-aided diagnosis in chest
radiography, Comput. Med. Imaging Graph. 31 (4–5) (2007) 212–223.
 [4] Q. Li, R.M. Nishikawa, Computer-aided Detection and Diagnosis in
Medical Imaging, Taylor & Francis, 2015.
References:

33.  [5] C. Qin, D. Yao, Y. Shi, Z. Song, Computer-aided detection in chest radiography based on artificial
intelligence: a survey, Biomed. Eng. Online 17 (1) (2018)
 [6] A.A. El-Solh, C.-B. Hsiao, S. Goodnough, J. Serghani, B.J. Grant, Predicting active pulmonary
tuberculosis using an artificial neural network, Chest 116 (4)(1999) 968–973.
 [7] O. Er, N. Yumusak, F. Temurtas, Chest diseases diagnosis using artificial neural networks, Expert Syst.
Appl. 37 (12) (2010) 7648–7655.
 [8] O. Er, C. Sertkaya, F. Temurtas, A.C. Tanrikulu, A comparative study on chronic obstructive pulmonary
and pneumonia diseases diagnosis using neural networks and artificial immune system, J. Med. Syst. 33 (6)
(2009) 485–492.
 [9] S. Khobragade, A. Tiwari, C. Patil, V. Narke, Automatic detection of major lung diseases using chest
radiographs and classification by feed-forward artificial neural network, in: 2016 IEEE 1st International
Conference on Power Electronics, Intelligent Control and Energy Systems (ICPEICES), IEEE, 2016,
 [10] G.G.R. Schramek, D. Stoevesandt, A. Reising, J.T. Kielstein, M. Hiss, H. Kielstein,
Imaging in anatomy: a comparison of imaging techniques in embalmed human cadavers, BMC
Med. Educ. 13 (1) (2013) 143.
References:

34. Thank you…!