Healthtech Disruption: An AI a Day Helps the Doctor a Long Way

The world is changing much faster than we can imagine, and this six-part series goes behind the scenes to understand how. So far, Indian women have been fighting a losing battle against breast cancer. More women in India succumb to breast cancer than any other form of cancer. The breast cancer rate in India is as high as 25.8 per 100,000 women, while all forms of cancer together range between 95 and 104. Around 1.8 million Indian women are expected to live with breast cancer by 2020. Worse still, India has a low survival rate for breast cancer, according to a study published in Lancet earlier this year. Only 66.1 percent of Indian women diagnosed with the disease between 2010 and 2014 are surviving, while US and Australia had survival rates as high as 90 percent for the women diagnosed with the ailment, during the same period.

Breast cancer experts attribute the low survival rate in India to cases being detected at the third or fourth stage where treatment is much more difficult. This happens because screening for breast cancer in Indian women is very low, and the widely used mammogram – an X-ray based imaging test–has its own limitations. Doctors typically advise women aged between 45 and 74 years, to get a mammogram done every two years as a precautionary measure. A mammogram test costs anywhere between ₹2000 and ₹4000 depending on the diagnostic center and the technology of X-ray machine used. But cost is just one of the many factors that discourage women from undergoing breast cancer screening. Other issues include accuracy (mammograms are more likely to miss cancers in women with dense breasts), pain during the test and cultural inhibitions. Use of X-rays is not considered safe as they cause radiation.

Also mammogram is not the most convenient test for women who have undergone mastectomy or surgical removal of one or both breasts, partially or completely. It is evident that India is losing the battle against breast cancer at the very first stage — diagnosis itself. But there is cause for hope, now that machines have evolved to the point that they can perform some tasks even better than humans. The adoption of artificial intelligence in healthcare is slowly on rise. At a basic level, chatbots help patients to book an appointment with a specialist doctors. Further, AI can also advise doctors on treatment plans, remind patients about their medicines, track their physical activity and monitor vital body parameters using internet of things. But one area that’s ripe for disruption is diagnostics — as successful diagnosis depends on recognition of patterns, mostly visual. That is where AI has an edge compared to humans.

The AI edge

Bengaluru-based health technology startup Nirmai has developed an AI-based kit to test for breast cancer. Founded by two former Xerox executives Geetha Manjunath and Nidhi Mathur — after three and half years of research and validation — Nirmai works with the help of artificial intelligence layered on top of thermal imaging. The kit is low-cost, non-invasive, portable, painless and radiation-free, addressing most of the problems that conventional mammograms grapple with. Nirmai, or Non-Invasive Risk Assessment with Machine Intelligence, uses digital infrared thermal camera to capture the images of the breasts and detect minute variations in heat patterns. Cancerous cells produce more heat compared to normal cells due to excessive formation of blood vessels and high rate of metabolism. In traditional diagnosis, the thermal images are scanned by a radiologist. But here is a problem — not all heat forming cells can be attributed to cancer and verification by a radiologist takes lot of time and effort. Also there are chances of false positive or missing the cancer altogether.

This is where automation comes into the picture

The back-end computers based on cloud are extensively trained on heaps of mammogram data using machine learning algorithms and cloud technologies. These computers then apply artificial intelligence to analyse and interpret data and will throw up results on whether the abnormal heat patterns are related to cancerous cells. Niramai’s screening device can detect tumors five times smaller than what a clinical exam can catch at a much lower cost than that for a conventional mammogram. “The cost will fall further as volumes pick up,” says Geetha Manjunath, CEO and CTO of Niramai. Manjunath says the product is slowly finding acceptance with doctors and hospitals as they compare the results of mammograms and thermolytics. Manjunath calls her product as thermolytics, and claims her product has 16 percent higher accuracy than mammography at one-tenth the price of mammogram machine. A mammogram machine costs around ₹1 crore and is not portable and needs a radiologist to operate. Given social and cultural reasons, more often women find it uncomfortable to get test done by a male radiologist. Manjunath said her company is in discussions with multinational device makers to distribute their solution across India and globally in future.

An AI pathologist

While Nirmai is trying to disrupt breast cancer screening, another Bengaluru-based AI startup SigTuple has built a solution that automates the job of a pathologist by studying microscopic images of blood samples, interpreting them and issuing reports. A pathologist is a doctor who specializes in interpreting laboratory tests by evaluating samples of body tissue to diagnose disease. For instance, in the most commonly-prescribed blood test or Complete Blood Count (CBC), a pathologist or technician has to view the slide under the microscope and manually count their number and analyze their morphology (structure) to find any abnormalities. CBC test is used to measure components of blood such as red blood cells, white blood cells, hemoglobin, plasma and platelets. A pathologist has to review every slide under the microscope before giving out the report. SigTuple’s solution automates that process — all it needs is a high-resolution image of blood sample and its AI technology does the rest.

“Since a human is reviewing the samples under a microscope – the outcome of the analysis is the function of his or her skill sets, experience and even the state of mind. That is the reason why we see lot of variation in reports from one lab to another,” said Rohit Kumar Pandey, CEO and co-founder of SigTuple.

“We have developed AI models which can itself do this entire analysis, and present virtual report which the pathologist using their handheld device can review and approve the report sitting anywhere in the globe,” Pandey said. SigTuple study found out that for every 100 manually-analyzed reports – pathologists differed on their conclusion in 33 percent of the cases. The error rates were sometimes as high as 60-70 percent, while mean average error for SigTuple AI solution is less than 3 percent, much below the acceptable range of 7-8 percent. It’s not just the accuracy alone – the AI solution unlocks the capacity of pathological laboratory by five times.

Typically, a pathologist takes 5-10 min to analyze a slide, at best he can review 100 slides in a day. The SigTuple solution now helps him to review at least 500 slides. Pandey says SigTuple is not a threat to pathologists but enables them to be more efficient and productive. “There are around 90,000 pathologists eligible to sign a report, and close to 300,000 path labs in the country – there is a clear shortage of medical experts and scaling of tests with existing resources is the need of the hour,” Pandey added. Founded in 2015 by three former American Express executives Pandey, Apruv Anand and Tathagato Rai Dastidar, SigTuple so far raised around USD 6 million from investors including the likes of Flipkart co-founders Sachin Bansal and Binny Bansal. The startup now looks to commercialize its AI solution.

Brining down medical costs

To be sure it’s not just Nirmai or SigTuple, startups like Ten3T (portable, easy-to-use electrocardiograms) and iNICU (child health monitoring), among others are developing disruptive diagnostic solutions that will significantly lower costs, while making physical distance a non-issue. They and others in this area do this via cloud-based linkages to hospitals and clinics, chatbots, smart apps and AI-enabled data analytics. “This means that sensors, real-time tracking and analytics will enable us to take pre-emptive charge of our own health, helping us to live more aware, healthier, longer lives,” said Manish Singhal founding partner of pi Ventures. Pi Ventures invests in early stage startups using artificial intelligence, machine learning and IoT to solve real world problems in healthcare. The VC firm has invested in Nirmai, SigTuple, and Ten3T.

Even big hospitals are now adopting artificial intelligence technologies in healthcare delivery. Manipal Hospitals for instance has been using IBM’s Watson – an artificial intelligence based computing platform to support its cancer specialists for last two years. Watson analyzes high volumes of data, understands complex questions posed in natural language, and proposes evidence-based answers on cancer treatment. Oncologists in India and across the globe are struggling to keep up with the large volume of research, medical records and clinical trials. Further, doctors face grueling battle to stay up to date about best practices in treatment and care management. “It’s early days — but there is a bright future for the AI,” says Ranjan Pai, the Chairman of Manipal Hospitals “IBM Watson done it for cancer care, there are few other areas where we are talking,” Pai said.

Pai indicated that digital pathology and radiology are other areas where he sees lot scope for artificial intelligence based application. “We look at it from doctor-patient perspective, its utility, cost benefit ratio, effectiveness the cost benefits and how does it work, before embarking on any artificial intelligence solution. India’s largest healthcare provider Apollo Hospitals isn’t far behind. Apollo partnered with Microsoft last month to develop and deploy new machine learning models to predict patient risk for heart disease and assists doctors on treatment plans. The team is already working on an artificial intelligence -powered Cardio API (application program interface) platform. Apollo also recently signed up with IBM to access Watson across its hospital network to aid its cancer specialists.

Access to data

Arup Roy, Research Vice President at research and consulting firm Gartner says the big challenge in India in terms of leveraging AI in health is having data in the digital format. “Hospitals need to keep medical history (in a digital format) for AI to be successful. The lack of proper sophistication in hospital information systems is a challenge,” Roy says. Pai concurs with Roy. “All these things work better if you have more and more data,” Pai says. His company now stores a lot of data on cloud which is accessible. With exception of a few large metros and tier-II cities – India has been facing acute shortage of healthcare infrastructure from hospitals to human resources like specialist doctors, radiologists and paramedics, while the costs of healthcare shoots up.

According to government data India has 0.6 doctors and 0.8 nurses per thousand of population, while the number of beds per thousand stands at 1.5. World Health Organization prescribes 1 doctor and 3 nurses per thousand population and 5 beds for 1000 people. “Essentially, what we need is to fill the gap between the needs of the plenty and the services of the few. In my view, artificial intelligence (AI) has the capability to enable solutions that form the critical middle layer of access—making healthcare accessible and affordable to a large population base at the same quality level irrespective of people’s social standing,” Singhal of Pi Ventures said. – Money Control