Unveiling COVID-19 from Chest X-ray with deep learning: a hurdles race with small data
Unveiling COVID-19 from Chest X-ray
with deep
learning: a hurdles race with small
data
Introduction:
COVID-19 virus has rapidly spread in mainland China and into multiple countries worldwide . Early diagnosis is a key element for proper treatment of the patients and prevention of the spread of the disease. Given the high tropism of COVID-19 for respiratory airways and lung epithelium, identification of lung involvement in infected patients can be relevant for treatment and monitoring of the disease.
Fig. 1: Example
Chest X-Ray images of: (a) non-COVID19 infection, and (b) COVID19 viral
infection
Figure 2: Example Chest X-Ray images from the
dataset, which comprises of 13,975 Chest X-Ray images across 13,870 patient
cases from five open access data repositories: (a) COVID-19 Image Data Collection,
(b) COVID-19 Chest X-Ray Dataset Initiative, (c) RSNA Pneumonia Detection
challenge dataset, (d) ActualMed COVID-19 Chest X-Ray Dataset Initiative, and
(e) COVID-19 radiography database.
Deep
Learning for chest x- Ray
Deep Convolutional Networks
(DCNNs) are being constructed to analyse chest images and diagnose common
thorax diseases and differentiate between viral pneumonia and non-viral
pneumonia. While many common viruses can cause pneumonia, the ones with viral
pneumonia cause substantial differences in X-Ray images. Which means that every
case of viral pneumonia will contain variable visual appearances. Moreover, finding
a dataset with positive samples poses another problem. Therefore, it is crucial
to develop a model which can overcome these pathological abnormalities and
detect the virus with high accuracy.
Fig. 3:
Original image (a) and extracted lung segmented image
How
does it work?
CNN has the
ability to learn automatically from domain-specific images and hence
differentiates itself from classical machine learning methods. Different
strategies can be implemented to train CNN architecture to acquire the desired
accuracy and results. In this paper, we have used a similar model of deep
convolutional neural network for the analysis of chest X-Rays. The collection
of medical data and reports is a difficult task. So, the dataset used is a
combination of five open-source datasets.
Conclusion
It was
proposed a deep convolutional neural network designed specifically for the
detection of COVID-19 cases by implementing computer vision and image analysis
on Chest X-Ray images gathered from five open access data repositories. The
experimental results show that the proposed model had the best performance
accuracy on the validation set. Further, it was investigated and applied
different model parameters in order to gain deeper insights on the Chest X-Ray
features critical for classifying Covid and non-Covid patients which can aid
clinicians in improved screening as well as improve trust and transparency.
Resources:
Comments
Post a Comment