Dr D Narayana, Professor, AIML, Chandan S, AIML Student, Jayshree V, AIML Student, Shardul S, AIML Student, Kishore S, AIML Student, Great Learning discuss how X-ray along with the assistance of AI can help detect the type of pneumonia, so that doctors can prioritise cases based on the level of urgency as affected nations might have extremely limited resources to test patients for COVID-19
COVID-19 is a viral respiratory illness caused by a coronavirus that has not been found in people before. The outbreak started in late 2019 and developed into a global pandemic by March 2020. The virus is called severe acute respiratory syndrome coronavirus 2 and it is abbreviated as SARS-CoV-2. In December 2019 Chinese authorities identified a cluster of similar cases of pneumonia in the city of Wuhan in China. The growth rate of the disease is exponential.
Why X-ray images are important:
One of the important ways to stop the spread is suspect person testing. Testing allows infected people to know that they are infected and helps them take measures to reduce the probability of infecting others. The most common tests are PCR tests. However, the affected nations might have extremely limited resources to test the patients. The extreme cases of COVID-19 patients develop COVID-19 induced pneumonia. Efforts are being made to evaluate if X-ray / CT images could be used to help diagnose the disease.
The advantages of chest X-rays include their low cost and easy diagnosis. When a patient walks in, X-ray is the easiest mechanism as he can get diagnosis reports in a few minutes. This process doesn’t cost much (approx. Rs 150 depending upon the location) compared to advanced techniques such as CT images. The risk of transmission is limited. If the AI based system can detect whether the patient has pneumonia followed by the type of the pneumonia, it can help the doctors prioritise the cases based on the level of urgency. X-ray machines are also portable hence they can be carried to the rural parts of the country with limited access to healthcare resources.
Our workflow is presented below:
This is an ensemble method which is robust with lesser deviation. We are able to get above 90 per cent accuracy for detecting healthy / unhealthy and different pneumonia cases.
Recommendation / Conclusion:
Using the proposed AI techniques based workflow we are in a position to analyse the patient’s condition using x-rays by spending minimal time of radiologists and patient’s money. These sort of solutions are needed for countries like India where the population size is huge.
Great attempt. Shouldn’t the last step in the workflow diagram be opposite? i.e. Covid induced pneumonia (viral infection) swap with non-covid induced pneumonia (bacterial infection).