The spread of new coronary pneumonia, the world has more than 204,000 people have been confirmed infected with the new coronavirus, the death toll of more than 8,200 people. Different people are responding to the outbreak in different ways. And computer scientists and machine learning researchers are responding to the epidemic in a way they know it is: compiling data sets and building algorithms to learn from them.
Google’s data science competition platform, Kaggle, already has a COVID-19 case data set and is updated daily. The data are reliable, including the patient’s age, location, when symptoms begin, time of exposure, time to enter the hospital, and so on. More than 400 people used the data in their analysis.
A researcher at the University of Montreal collected and published a database that included dozens of CT scans and chest X-rays. These images were taken from open disease studies.
Johns Hopkins University has created an impressive dashboard that contains reliable data from sources and is regularly updated to show the world the spread of disease and mortality. Code that can be copied and modified is available on GitHub.
Other data sets come directly from hospitals treating patients, which have quickly tried to change machine learning models to help doctors find signs of disease.
Here are some of these papers:
Paper title: Deep learning quantitative study of COVID-19 lung infection in CT images
Shanghai researchers designed a system that, along with manual checks, could reduce the time it takes to analyze CT images from hours to about four minutes.
Paper title: Coronary virus (COVID-19) pandemic fast AI development cycle: the use of deep learning CT image analysis for automatic detection and initial results of patient monitoring
The paper also claims to detect the presence of COVID-19, and to visualize the effects of the virus on the lungs to track disease progression over time.
Title: Abnormal Breathing Pattern Classifiers May Help To Screen COVID-19 Infections on an Accurate and Unobstructed Scale
Here, the researchers looked for an auditory method to screen COVID-19 by analyzing a person’s breathing rate. The study is inconclusive, but it is a new way to test the virus in less invasive ways.
Paper Title: Deep Learning System Screening 2019 New Coronary Pneumonia
The work attempted to distinguish between pneumonia in PATIENTS WITH COVID-19 from common influenza.
Address of the paper:
Title: Predicting the Criticality of Severe Covid-19 Infection Patients using the following three clinical characteristics: Clinical Data Prediction Model based on Machine Learning in Wuhan
Using electronic health records from nearly 3,000 patients in Wuhan, the researchers built an algorithm that predicted the accuracy of the survival rate of patients with severe illness estowasto.