In April 2020, when the Covid-19 pandemic swept Mumbai’s crammed coastal belt, there was panic across the city. The Brihanmumbai Municipal Corporation (BMC) was quick to call Dr. Prashant Warier, the founder and CEO of Qure.ai for help. A developer of cloud-based artificial intelligence (A.I.) tools for medical imaging, the four-year-old start-up was already sitting on five patents and had raised $16 million in a series-A funding. BMC wanted to check if Warier’s new A.I. solutions for chest X-rays could detect the Covid-19 virus.

Dr. Hemant Deshmukh, dean of KEM Hospital and head of Covid-19 taskforce in Mumbai, points out how the administration faced a huge challenge initially as the number of RT-PCR tests daily was abysmally low and that the results were available only after 24 hours. Qure.ai offered a unique Covid-19 solution to check the presence of the virus in X-rays. Their tool, called qXR, was built into the X-ray machine which sent the anonymised data (no patient’s data) to the cloud. In the absence of a digital X-ray, it used the mobile app on android to click a picture of the analogue X-ray from a view box. The tool could interpret the X-rays quickly and report positive cases, mostly in a minute. The A.I. tool predicted the Covid-19 likelihood—high, medium, low, and none. If the X-ray showed a possibility with high and medium likelihood, they would collect the swab immediately and send it for Covid-19 confirmation test.

“We realised the importance of uploading the chest X-rays on the A.I. platform from wherever the patient population was large. The results were amazing. The patients, who were suspects, were immediately put on treatment for the simple reason that the X-ray showed findings of Covid-19. Even these patients were triaged and you will be amazed to know that the RT-PCR results on them were almost 80-85% positive. We started chasing the virus by deploying portable digital X-ray machines,” says Deshmukh, adding that A.I. also helped them identify hotbeds in the city quickly.

Mumbai was doing a lot of screening for Covid-19, with a mobile bus stacked with an X-ray machine, going around the city, especially in Worli, Koliwada, Dharavi, and other areas that were reporting a surge in the pandemic.

Yet again, the necessity turned out to be the mother of a new invention.

Though they discontinued the Covid-19 bus, the qXR tool is operational in at least a dozen hospitals in the city today. “We get around 300-400 X-Rays a day for reporting,” says Warier. His company has deployed the Covid-19 algorithms in several hospitals run by the National Health Service (NHS) in the U.K., and various medical institutions and hospitals in Mexico, the U.S., and Italy.

“A.I. solutions are faster, accurate and cost-effective,” adds Warier, an IITian from Delhi who went to Georgia Institute of Technology in the U.S. for his MS and Ph.D.

Health worker in a primary care centre using Qure's software
Health worker in a primary care centre using Qure's software
Image : Courtesy of Qure.ai

Changing healthcare

The other marquee product developed by the Mumbai-based start-up is the qER tool which can detect intracranial haemorrhages, subdural bleeds, midline shift, and mass effect (for tumour detection) to offer quick and accurate head CT scan analysis. Nearly half of all stroke related deaths are due to intracranial bleeds/haemorrhagic stroke. Qure’s product qER can quickly detect intracranial bleeds in under a minute with a high accuracy comparable to that of an expert radiologist. Medica, one of the largest UK teleradiology firms, and U.S. firms such as vRad are making use of qER.

Another application of the Qure.ai technology is in the area of remote medical diagnostics, paving the way for patient treatment in under-developed and developing countries. Their A.I. tools can accurately diagnose TB, a disease that claims millions of lives annually. Qure.ai has achieved unprecedented success with its mobile medical van programme that deploys onsite, cloud-based TB screenings in under-served regions of the world, where advanced in-person diagnostics are scarcely available.

“We can detect TB much better than a radiologist,” says Warier. He is quick to add that their tools are to supplement the services of radiologists, and not to replace them.

In January 2021, Qure.ai got an endorsement from the World Health Organisation (WHO) for using A.I. as an alternative to a human reader.

At present, Qure.ai is spread over 35 countries, mostly in Asia, Middle East, Africa, Europe, LatAm, the U.K. and the U.S. In countries like India, Africa, and South-East Asia, its A.I.-assisted medical imaging solutions are already helping small medical teams to turn around lifesaving diagnoses in a fraction of the time normally required. “In the Philippines, X-ray readers are deployed in mobile vans. Typically, they would have taken at least five weeks to read the X-rays whereas we read it in one minute. This is what we are doing in many TB screening countries. TB is a poor man’s disease. It is found in developing countries and in poorer pockets where it is even more difficult to get the x-rays read,” he says.

Journey

For 41-year-old Warier, the journey from adtech to the niche, A.I.-assisted healthcare sector wasn’t accidental.

In 2012, he was running a startup in Mumbai called Imagna Analytics, an adtech company that worked with ecommerce giants, focusing on the advertising technology space. It used cookie data to figure customer behaviour patterns and target advertising. “Those days, companies were still figuring out how to personalise their ads. I was using A.I. to target the right ads to the customer, based on your previous behaviours, purchases, and previous click history,” he says.

In 2015, Imagna was acquired by Fractal, a company that provided analytical consulting services to Fortune 500 companies. He joined Fractal as chief data scientist.

“At Fractal, I started incubating Qure, rather than just providing services.” In fact, Fractal, an analytics consulting firm, provided the seed fund for Qure.ai. Armed with A.I., Warier stepped into the vast healthcare imaging space that includes radiology and pathology.

Can a machine understand radiology and pathology like the way it understands regular images?

Six years ago, image recognition was still at a nascent stage. “I realised A.I. and deep learning could do extremely well on image recognition even better than human beings. We looked at radiology and pathology imaging to figure out if we can apply the A.I. tools on it. We did a lot of research. In radiology, there is a lot of historical data. In hospitals, they store about seven years of history. I knew if we could get anonymous access to this data, we could train our algorithms on it,” says a confident Warier.

On the contrary, he found pathology imaging very ad-hoc in a digitalised process, with little historical data.

“A.I. requires a lot of data to train. And we thought the radiology department is an area where we could get access to historical data. That is how we started focusing more on radiology imaging. We spoke to a lot of radiologists around the world, from the U.K., the U.S., Middle East, and India. About 1 billion-plus X-rays are taken annually around the world. And most radiologists do not want to read X-rays because the cost of reading it is very low, at around ₹15-₹20.”

Hence, most of the time X-rays are read by technicians or by general practitioners in India.

In the U.K., they are read by radiographers, and not by radiologists. The NHS in the U.K. has a huge backlog of X-rays, according to Warier. “The reports of X-rays that were taken a month ago are still not ready. This is what happens in the U.S. too. To add to that, the error in reporting in the US is substantial, at around 20-23%. Significant abnormalities are missed on an X-ray because people are not paying attention. The pricing is similar in the US too, $2 or lower for reading an X-ray whereas you get paid $50 or more for reading a CT or MRI scan. It can also go up to $100-200 depending on different modalities.”

For Qure.ai, chest X-rays presented a huge opportunity. From X-ray reporting, it expanded its horizon to the entire gamut of pulmonary-related conditions such as TB, Covid-19, lung cancer, and others.

Scaling up

Steadily, Qure.ai has entered into the head CT scan space too. This is a type of scan that requires a fast reading. Getting a CT done takes only a few minutes, but reading it could take hours. If you are looking for a tumour or atrophy, you can do an MRI. You do a CT in case of a trauma or stroke which needs immediate intervention. A delay in reading a CT can be fatal for the patient. So, this is where A.I. comes to the help especially when there is a huge volume. Processing a lot of data at the same time is a Herculean Task for humans. “But using A.I., I can process innumerable X-rays or CT scans if we have adequate computer resources to do that. A.I. can process the CT scan and provide a report within a minute on the mobile phones of the radiologist,” says Warier.

At the government medical college in Thiruvananthapuram, Kerala’s capital, Qure.ai has implemented its A.I. algorithm on the CT scanner. Every time it gets a CT scan with a critical abnormality, it immediately alerts radiologists on their mobile phone with a telegram message. Similarly, it has deployed the solutions at Baptist Christian Hospital Tezpur in Assam, just a month ago.

“This product has been cleared by the U.S. FDA,” says a proud Warier. It made Qure.ai the first Indian A.I. company to get an FDA approval.

For his company, another focus area has been lung cancer. It recently signed with AstraZeneca for undertaking lung cancer screening for the pharma major in various parts of the world. “We may have taken an X-ray for an immediate critical condition like pneumonia, some type of cough, or some other infectious symptoms. While focusing on immediate conditions, radiologists may miss out on lung modules that are indicative of cancer. An early detection can lead to better chances of surviving lung cancer. So, A.I. can alert the radiologist who can follow it up with a CT or biopsy,” says Warier.

More importantly, in case of a stroke or trauma, the A.I. tool can identify the bleed source and quantify it. This can help direct the patient’s clinical pathway between trauma and stroke. The best way to treat it is to give anticoagulants that will quickly dissolve the clot in the brain. But you would want to ensure that there is no bleeding in the first place. If there is one, you can’t give anticoagulants as the patient will bleed more. A.I. can advise the radiologist if he can prescribe an anticoagulant, without any time being wasted in the process. The other important factor is that the hospitals that are situated in remote areas may not have access to quality radiologists.

“Our focus has been on four broad markets: all products in the U.K., the U.S., Middle East, and the TB screening market globally,” says Warier.

Qure employs 70 people today, most of them being doctors, IT specialists, and data scientists, in five countries—the U.S., the U.K., Middle East, Germany, and India. The company is split into multiple parts: the R&D team that builds the algorithms; the engineering team which converts algorithms into solutions that customers can use; the product team that talks to customers and figures out the product roadmap; sales team, regulatory team, and the client success team.

The management of Qure.ai is convinced of its commercial strategy. “We do not get paid adequately in India and it is difficult for us to make it commercially viable here. We have found a big market outside India,” Warier says.

The company has two revenue models—per-scan basis and subscription basis. Under the subscription scheme, one can use the A.I. tool for an unlimited number of scans. There are customers using both the schemes, depending on the volume that they handle. Some customers have huge volumes in which case Qure.ai levies a fixed annual price. The tools are open to use by doctors or radiologists.

Follow us on Facebook, Twitter, YouTube & Instagram to never miss an update from Fortune India. To buy a copy, visit Amazon.