Indian researchers develop AI that may help in the fight against Covid-19

Given limited testing facilities for Covid-19, there's a rush to develop AI tools for quick analysis using X-rays. Experts say it can be very useful in smaller towns where access to CT scans is poor

India could see a staggering one million deaths from Covid-19 by August 1, according to an editorial in the British medical journal Lancet. There have been at least 22,992,517 confirmed cases of coronavirus in India and about 249,992 people had died. Experts say that the actual death count far exceeds official figures. 

With over 400,000 cases of Covid-19 every day, India is in dire need of a viable detection solution. Identifying and isolating suspected patients quickly has become critical to controlling the spread of coronavirus. Early detection and diagnosis of Covid-19 can help in saving lives. However India is facing a huge challenge of testing processes and manpower.

To combat covid fatalities involving the lungs, the Defence Research and Development Organisation’s (DRDO's) Centre for Artificial Intelligence and Robotics (CAIR), with the support of 5C Network and HCG Academics has developed ‘Atman AI’, an artificial intelligence algorithm that can detect the presence of Covid-19 disease in chest X-rays. Atman AI is used for chest X-ray screening as a triaging tool in Covid-19 diagnosis, a method for rapid identification and assessment of lung involvement. This will be utilised by online diagnostic startup 5C Network, India’s largest digital network of radiologists, with the support of HCG Academics.

Dr U K Singh, director, CAIR, DRDO said the development of Atman diagnostic tool is a part of DRDO’s effort to help clinicians and partners on the front line to have the tools they need to rapidly diagnose and effectively treat Covid-19 patients. 

“Given the limited testing facilities for coronavirus, there is a rush to develop AI tools for quick analysis using X-rays,” said Singh. “The tool will help in automatically detecting radiological findings indicative of Covid-19 in seconds, enabling physicians and radiologists to more effectively triage the cases, especially in an emergency environment.”

Experts say that triaging Covid suspect patients using X-ray is fast, cost-effective and efficient. It can be a very useful tool especially in smaller towns in the country owing to the lack of easy access to CT (computed tomography) scans. This will also reduce the existing burden on radiologists and make CT machines which are being used for Covid be used for other diseases and illness owing to overload for CT scans. 

The average cost of a CT chest scan today ranges between Rs 1,500-3,500. But the average price of X-ray chest scans ranges between Rs 300-600, thereby indicating that Atman ensures a significant cost reduction capability.

5C Network has been doing over 5,000 covid scans every day. It said the AI-powered algorithms have proven to help radiologists triage patients better.

“We are excited to develop Atman AI for Covid detection in chest X-rays,” said Kalyan Sivasailam, co-founder and CEO,5C Network. “Utilising the algorithms for chest X-ray is an effective triaging tool which can be accessible to the common man in the remotest districts of this country. This will have a significant impact on timely care and appropriate treatment.”

Siddhartha Hospital at Kakinada in Andhra Pradesh is using Atman AI through 5C Network. This has allowed us to screen more patients for covid than before. Radial Diagnostics Bangalore in Karnataka is also using the innovation through 5C. This is allowing more patients to have access to covid diagnosis as it is much cheaper than CT Thorax scans which were being used before.

The AI feature with existing models have improved the accuracy of the software. Being a machine learning tool, the accuracy will improve continually. Chest X-rays of RT-PCR (reverse transcription polymerase chain reaction) positive hospitalized patients in various stages of disease involvement were retrospectively analyzed using AI (deep learning and convolutional neural network) models by an indigenously developed deep learning application by CAIR-DRDO for Covid-19 screening using digital chest X-rays. The companies said the algorithm showed an accuracy of 96.73 per cent.

According to Dr Vishal Rao, dean academics, centre of academic research, HCG Cancer Hospital, the innovation will help to reassure vulnerable health care workers. It would improve efficacy in hospitals without increasing the financial burden for patients and healthcare systems.  “Furthermore, similar methods can be used to assess predominant respiratory diseases affecting the vulnerable population and programs initiated.”

Started by Kalyan Sivasailam and Syed Ahmed in 2016, Bengaluru-based 5C Network is a digital platform that enables the storing, sharing and interpreting of radiology images across India. The firm said it helps hospitals and diagnostic centres maximise the ROI (return on investment) on their radiology machines by making specialists available 24x7. 5C has reported over 2.5 million scans across India.

DRDO and 5C Network are not the only organisations using AI to fight the pandemic. Qure.ai, a healthcare startup backed by Fractal Analytics has re-purposed its chest X-ray AI tool to detect signs of Covid-19. Mumbai-based Qure.ai has developed AI-powered virtual care solutions that can identify people at high risk of Covid-19, much before they reach the hospital emergency rooms. The platforms help track, manage and prioritise the testing and improve diagnosis of Covid-19. Srikanth Velamakanni, co-founder, group chief executive, and executive vice-chairman of Fractal Analytics, recently said that multiple countries are using Qure.ai’s solutions including X-ray for covid detection.

Researchers, healthcare providers, and many others around the world are still grappling with Covid-19. It remains challenging for doctors to predict how a patient’s condition may change over the course of the disease. With resources under unprecedented strain, it’s important that hospitals know whether patients are likely to need escalated treatment and plan accordingly.

Facebook’s model using sequential chest X-rays can predict up to four days (96 hours) in advance if a patient may need more intensive care solutions, generally outperforming predictions by human experts. These predictions could help doctors avoid sending at-risk patients home too soon, and help hospitals better predict demand for supplemental oxygen and other limited resources.

GE Healthcare has also launched a new AI algorithm to help clinicians assess the correct placement of ventilator tubes. This is a necessary and important step when ventilating critically ill Covid-19 patients. The solution is embedded on a mobile x-ray device for automated measurements, case prioritization and quality control. Research shows that up to 25 per cent of patients intubated outside of the operating room have misplaced ETTs (Endotracheal Tube) on chest x-rays. This can lead to severe complications for patients, including hyperinflation, pneumothorax, cardiac arrest and death.

This year tech company Microsoft said in a blog post that a digital diagnostic tool that uses AI and cloud computing to accurately read vast numbers of chest X-rays – faster than a radiologist can – is helping doctors identify, triage and monitor Covid-19 patients.

South Korean software company Lunit’s algorithms have been trained to read X-rays and originally could detect signs of 10 major chest diseases. These include cancers, with an accuracy rate of 97 per cent to 99 per cent. When the pandemic struck, its developers quickly adapted them to also scan for signs of Covid-19 – including pneumonia, which is often present in infected patients.

Using the cloud computing power of Microsoft Azure, Lunit’s technology generates location information of detected lesions in the form of heatmaps, according to the Microsoft blog post. It also makes abnormality scores that reflect the probability that the detected lesion is abnormal and needs further investigation by a radiologist.