8 Ways How Computer Vision Has Been a Sentinel Against COVID 19

2020 was a challenging year for the entire world, as humankind continued to battle against the deadly COVID-19Disease.And the fight hasn’t ended yet. But it certainly has been made easier, thanks to Computer Vision. A subfield of AI, Computer Vision has simplified several complex problems in the healthcare industry, thus helping in controlling COVID-19.
The biggest challenge in containingCOVID-19 has been its infectious nature and the faint trail that the virus leaves. The virus outbreak left governments, and medical communities of the most developed countries perplexed, threatening the entire human population in the process.
As researchers around the world worked in unison to explore new diagnostic techniques to battle COVID-19, Computer Vision came as a breakthrough. With the perfect blend of deep learning and machine learning, Computer Vision made disease diagnosis, prognosis, medical image analysis, prevention, control, and drug discovery a lot easier for the healthcare and medical sector. Wondering how? Let us delve into the eight best applications of Computer Vision in fighting the COVID-19 crisis.
8 Applications of Computer Vision in COVID-19 Diagnosis
To contain COVID 19, the world needed a diagnostic process that is fast, reliable, and widely available. This is where Computer Vision came in to be the saviour! Here’s how CV collectively saved the world:
1.X-Ray Radiography
Thanks to Computer Vision, millions of people were saved due to the cost-effective digital chest x-ray radiography used to detect chest pathology. Here are some more insights into how CV has changed the face of X-Ray radiography:
COVID-Net is a deep learning computer vision model developed by Darwin AI from Canada to detect the traces of COVID-19 using chest radiography images.
The model was based on a dataset called COVIDx, which included 16,756 chest x-ray images obtained from 13,645 patients with symptoms of COVID.
This particular computer vision model has an accuracy rate of 92.4% and has been used worldwide for COVID-19 diagnosis.
The current COVIDx dataset is used as an archive of datasets in assignment help front-line medical workers determine which treatment to usefor an infection depending on the cause.
2.Computed Tomography
Computed Tomography (CT) is a test used to generate a detailed image of a patient’s chest using radiology. CT differs from an X-Ray as a CT scan lets you view the position of bones, muscles, fats, and organs as well. CT features such as consolidation affecting the lower lobes and ground-glass opacification showed observable symptoms of COVID-19 pneumonia. Here are some more features of Computer Tomography:
CT uses image segmentation with deep learning aims to identify the infected area of a patient.
In China, the Renmin Hospital of Wuhan University used a method based on the UNet++ semantic segmentation model to simplify and quicken the diagnosis process for patients who contracted COVID-19.
The CT model used by Renmin is based on a dataset of 46000+ images – a collection of both infected and healthy patients labelled by expert radiologists.
The dataset was created after studying 106 patients, out of which 51 patients were confirmed COVID-19 patients, and 55 of them were controlled.
The Wuhan model had an accuracy rate of 95.24% per patient, gaining a positive predictive value of 84.62%.
3.Masked Face Recognition
To limit and contain the spread of the virus, using masks and other protective gear became a norm. However, with masks on, the technology of face recognition hit a dead-end. This is again where Computer Vision systems stepped in to remove the glitch and facilitate the implementation of face recognition.
Based on a multi-granularity, a model masked for masked face recognition was created. The model was based on datasets of several masked face images and achieved an accuracy rate of 95%. The data contained 3 types of datasets of masked faces. Here’s more about the model:
MFDD or Masked Face Detection Dataset is used for the masked face detection model.
RMFRD, or the Real-world Masked Face Recognition Dataset, was a compilation of more than 5,000 masked pictures of 525 people and 90,000 unmasked images of the same 525 people.
SMFRD or the Simulated Masked Face Recognition Dataset is a simulated dataset of 500,000 masked face images of 10,000 people. 
4.Thermography
Infrared thermography is used as a strategy for early detection of people who are infected with the COVID-19 virus, primarily in public places such as malls, stations, and airports. Medical machines were created where infrared thermography was used to screen fever.
The smaller version of this approach is used in a mobile platform for screening automatic fever based on the infrared temperature of a person’s forehead.
Several organisations have started integrating infection screening in thermography and CCD cameras to measure vitals and limiting contact at the same time. This has been made possible using feature matching and MUSIC algorithms.
5.Disease Progression Score
Computer Vision has been used to develop better clinical management practices that could classify patients based on how severe their symptoms are. Computer Vision is also used to identify critically ill patients for prioritised medical attention using critical patient screening.
The model uses a disease progression score to classify infected patients. The score is a total of the infected areas as gathered from the CT images, and this score is then used to identify the most critically ill patients and measure their progression. Some important pointers to remember in this case:
Deep learning along with depth cameras are used to classify abnormal respiratory patterns.
The application enables an accurate and unobtrusive large-scale screening of people infected with the COVID-19 virus.
The application also uses the Respiratory Simulation Model (RSM) to bridge the gap between a large amount of training data and scarce real-world data.
People infected with COVID-19 are found to have more rapid respiration. Hence, the model is based on Gated recurrent units (GRUs) neural network to classify six clinically significant respiratory patterns to identify critically ill patients. The model can classify respiratory patterns with an accuracy of 94.5%.
6.Pandemic Drones
Pandemic drones have been created for remote sensing and digital imagery, which help in identifying infected people. Hence, the drones are used for prevention and disease management purposes. Such Computer Vision applications are also being used to monitor the lives of infected patients remotely and reduce contact. Another similar application used in disease management during the pandemic is vision-guided robot control for better 3D object recognition.
7.Germ Screening
Since COVID 19 is a lot about leaving trails of germs (the virus, in this case), germ scanning has been exceptionally helpful in combating COVID-19. To date, there is no definite treatment for the COVID 19 virus. But with a convolutional neural network, germ scanning can be used to identify bacteria using light-sheet microscopy images. What’s more, the model has already achieved over 90% accuracy. The focus of germ screening is on treating symptoms to improve a patient’s clinical condition.
8.Support Vaccination Development
Deep learning, especially feature representation learning, is now being used to analyse quantitative structure-activity relationship (QSAR). The model incorporates 360° images of molecular conformations into datasets for further deep learning. QSAR analysis then uses deep learning on a novel molecular image input technique, which can be used for discovering a potent drug that can be developed into a successful vaccine.
Parting thoughts,
Computer Vision technology has a multidisciplinary nature, making it potent enough to support the healthcare and medical industries in combating the challenges of the pandemic. As AI vision approaches become more popular, we can expect better applications to end the suffering that the whole world is living through at the moment.