Detection of COVID-19 From X-ray images Using Deep Convolutional Neural Network(CNN).
Coronavirus disease 2019 (COVID-19) is an infectious disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2).The disease was first identified in December 2019 in Wuhan, the capital of China's Hubei province, and has since spread globally, resulting in the ongoing 2019–20 coronavirus pandemic.About 3 million people have been suffering from this pandemic .Among several ways for detection, I tried two approaches using X-ray and CT-scan. I suppose using them is a faster, easier and less harmful method than others.Lungs of those patients(infected from Covid) were presented with patchy ground glass opacities(GGO),crazy paving appearances and air space consolidation.
This is a standard Convolutional neural network model for detecting covid-19 from X-ray.
Datasets are collected from several sources. Two categories of data were collected for training and testing:x-ray and CT-Scan. X-ray non covid(normal and pneumonia) data are extracted through kaggle whereas X-ray covid-19 datasets are extracted from the github open source. CT-Scan datasets ,both covid as well non covid were fetched from github open source.
| Non-covid image | covid image |
|---|---|
 |
 |
Due to less number of covid-19 positive images, images were augmentated by using the operations of horizontal and vertical flipping and rotation.
Several standard models like VGG16,ResNet50,InceptionV3 have been implemented for transfer learning.Among them, VGG-16 gave the best result. Due to some limitation of data (Noise and small) different methods have to be followed.
At first, X-ray data was trained in a standard pretrained model by freezing some layers.Almost 99% percent training accuracy and 98% of validation accuracy was achieved.After that CT-scan datas are also trained on the pretrained model of x rays.
Confusion matrix as well precesson table is shown below.
model can be downloaded from: https://drive.google.com/file/d/1VBgyJiurgMJCYz7AhCVhalZnP5XijulQ/view?usp=sharing
The model for CT-scan is still to be improved.The models can be further improved after availability of more covid-19 positive x-ray and CT-scans images.