Artificial Intelligence to deal with COVID-19 in Sri Lanka

Sampath Godamunne, BSc.(Eng.) MSc,

sampath@microrsg.com

Covid-19 is a deadly virus as of our current knowledge and experience. The dynamics of the virus is still may be a mystery for most of us. There exist a way to deal with this deadly microbe with the help of modern day knowledge. It’s nothing but mathematics, yet we prefer to call it AI or Artificial Intelligence. First we should look at the common knowledge of this virus as of WHO and other experts. They say even around 80-85% who contract Covid-19, shows mild symptoms and this asymptomatic nature of the disease, becoming a huge danger and carry the virus secretly among the community.  If we get back to our AI basics, the current knowledge on disease identification Machine Learning models provide a better way to deal with this Covid-19. For example, cancer identification Machine Learning models can identify breast cancers with very high accuracy through feature data sets. If we get back to our Covid-19, according to the presently available information, 80-85 % shows mild or less symptoms and balance 15-20% subjected to severe conditions and around 5-6% are becoming critical.

How we deal with Covid-19 with the help of AI (Artificial Intelligence). We must build a feature (or say characteristics/attribute) data set of these corona positive patients who are being treated at IDH and other hospitals. For example;  age, gender, blood group, early medical history, chronic illnesses, infected with SARs, etc. The decision of feature data set is very critical and should be done by the medical experts. I am sure they must be having a superior knowledge on what data to be taken for the analysis. There may be certain data sets currently available from china and Europe or even from WHO.  I am not sure whether they share those data to us rather than to big Phamas. Due to the mysterious behavior of virus, internationally available data sets may not be very well fit into Sri Lankan community. The data of currently Covid-19 positive patients will be very much vital when we build a Machine Learning model for Covid-19 of Sri Lanka. Those data sets could be categorized into desired categories like Mild or no symptoms, Severe, Critical etc. to train the model. Same time the set of actual data can be used to verify the ML Model we are building.

As we go on, with more patient feature data sets, this model can be further be strengthened to provide more accurate predictions on what community  is more susceptible or vulnerable to Covid-19.

Next question is what we can do with this knowledge. By doing so, there are numerous possibilities presented into our hands. This model will then help us to predict the most vulnerable community who will be critically ill, if they contracted with Covid-19. In order to find this community we should set up a system to get data from the public through a proper system via internet.

According to Wikipedia, here is the mobile community of Sri Lanka. 

This shows that mass data collection from public is not a difficulty as far as we have a good platform to do it. Once we collect data from public, we can run them through the Machine Learning Model, and predict & identify  which ones are more susceptible to Covid-19.

If we go back to the current knowledge of Covid19, it is shows mild symptoms for 80-85% and 15% become severely ill and 5-6 % will be critical. As of 23^nd March 2020, fatality rates are as follows from countries with more than 1000 positive cases. Even the hard hit Italy is currently having less than 10% of fatality rate. 

What illustrate here is, not everyone contracted with covid19, will be critically ill but there exist a community who are more vulnerable. As a country, we build this knowledge without delay, we can surely take all the possible measures to identify and save their lives easily which should be the common goal of our country at this time. With the limited medical supplies to treat the critically ill patients, Sri Lanka should take all the measures as a country to identify and keep these more susceptible community away from Covid-19. Then we can maintain the crisis without overwhelming demand for critical life support systems which is not widely available in our country.

Currently available technologies enable us to run all Machine Learning system at the sever level while collecting data via web. From ML modeling, mass data collection, isolation, monitoring could easily be managed with presently available technology stack which has already been built by human on earth with multi billion dollars or whatever currencies.

Use of right technology & human resource at the right time is the winning strategy. I strongly believe Sri Lanka has all the capacity and courageous to face & overcome Covid-19 crisis.  If anyone interested we can contribute.