Microsoft AI Network for Cardiology with Apollo Hospitals to bring new insights in predicting population-based heart diseases

“Nearly three million heart attacks happen in India every year and there are 30 million people who are suffering from coronary diseases. It is almost an epidemic in this country,” says Dr. J. Shiv Kumar, Chief of Cardiology, Apollo Hospitals, one of India’s largest private healthcare companies in India.

Despite the sheer number of cases, doctors in India are unable to identify the probability of cardiac ailments when patients come for their regular health check-up. While there are some scores or algorithms available worldwide that predict the probability of a patient having a heart attack in the next 10 to 20 years, doctors can’t extrapolate the same risk-factors and apply them to patients in India since most of them are derived from western studies and don’t have a high degree of accuracy for the Indian population.

“LDL cholesterol, for instance, which is high in western populations and one of the major causes of heart attacks, isn’t that prevalent in Indian populations and despite that they had heart attacks. Clearly, the risk factors are different in the two populations and we must figure out the emerging risk factors in Indian population,” Dr. Kumar explains.

Apollo Hospitals partnered with Microsoft’s AI Network for Healthcare to develop an India-specific heart risk score and better predict cardiac diseases for general population with the help of Apollo’s database and expertise in the field, and Microsoft’s cloud and AI tools. It is part of Microsoft Healthcare NExT, which aims to accelerate healthcare innovation through AI and cloud computing.

Mining data for the smoking gun

The team at Apollo Hospitals collected more than seven years of data from master health check-ups conducted at its hospitals across the country from 2010 to 2017, which consisted clinical and lab data of 400,000 patients. However, having data is one thing but mining and getting actionable insights out of it was going to be a challenge.

“This was a large transactional sample of data and we had to work with clinical experts and data scientists to ensure that the entire dataset, and not just a sample dataset, was correct,” explains Arvind Sivaramakrishnan, Chief Information Officer, Apollo Hospitals.

“It required significant amount of compute power and design guidelines from a technology perspective. This is where we used Microsoft Azure to upload the data to the cloud and some of SQL tools to put it into the data warehouse from where data scientists and clinicians could statistically correlate and train the machine learning models,” he adds.

The team also ensured that the security and privacy aspects of the data that was uploaded. They stripped off all patient identifiable information from the data sets and ensured the processes were in line with Indian IT laws as well as HIPAA and ISO 20071 guidelines for data privacy and security before the data was shared.

A team of data scientists and clinicians working on the data found that nearly 60,000 patients had been admitted to a hospital for a cardiac event after undergoing health check-ups. Of these, approximately 34,000 patients had two normal check-ups in a span of five years and yet had a cardiac event. Over 7,000 others had a cardiac event preceding one, two or three years of a normal health check-up.

Why did the health check-ups not show any signs of the impending cardiac event? Or were the signs there in the data but no one was looking at them? The team had found the smoking gun!

Building a heart risk score for India

The team fed the data of these patients into the machine learning model to find out the risk factors that caused these patients to have a heart attack despite having a normal health check-up. It was clear that there were different risk-factors, which current models were either not considering or not giving enough weightage in building the risk-factor score.

They worked with Azure Machine Learning services, which is an end-to-end service that helps accelerate the rate of Machine Learning experimentation by rapid prototyping, scaling up on virtual machines and proactively managing model performance. The Azure services include Libraries for Apache Spark and Visual Studio Code Tools for AI.

The team started with 100 health check-up risk-factors and 200 lab data points and correlated each factor in relation to its significance to the occurrence of the disease. Eventually, they narrowed down to 21 risk factors to build model to predict heart risk for Indian population.

But was it any better than the existing models?

A game changer

“We found that the accuracy of our model for Indian population in predicting the probability of coronary disease in the future was twice than the existing ones. It’s a game changer!” exclaims Dr. Kumar.

With the new heart risk score for India, Apollo Hospitals is looking at redefining how preventive health check-ups are done across its hospitals. The models help gauge a patient’s risk for heart disease and provide rich insights to doctors on treatment plans, assist early diagnosis and empower doctors with predictive solutions.

“This is just the beginning. It has given us a glimpse of how we can implement AI to get predictive healthcare. We could possibly apply it to other sub-specialities of heart problems,” says Dr. Kumar.

The team is already working on an AI-powered Cardio API platform, which would let anyone come to the hospital and be able to find their heart risk score without a detailed health check-up.

“All we need to do is fill in some relevant information into the AI system and we’d be able to classify them under mild, moderate or severe risk category. We’d be able to find this even before a detailed health check-up,” says Dr. Kumar.

From an operating process point of view, Sivaramakrishnan’s team has integrated the model into the Electronic Medical Records (EMR) for patients. “Every general physician who uses EMR for a patient undergoing a preventive health screening is able to access that data, get the risk score, and take action accordingly,” he explains.

“This collaboration marks a paradigm shift for patients and the management of non-communicable diseases the world over. Cardiac disease is amongst India’s leading causes of mortality and morbidity, which is exacting a toll on the health and well-being of our citizens. Given our large clinical database and significant pool of clinical talent, this partnership will help impact the global burden of cardiac disease,” says Sangita Reddy, Joint Managing Director, Apollo Hospitals. “This collaboration with Microsoft’s path-breaking technologies like Artificial Intelligence and Machine Learning will help better predict, prevent and manage heart disease in the country. While we commence this in India, we will validate the algorithms and work towards creating a global consortium to tackle multiple conditions in cardiovascular disease.”

Microsoft’s AI Network for Healthcare is an expansion of the Microsoft Intelligent Network for Eyecare (MINE), which is a global consortium of eyecare providers, research and academic institutions joining hands to apply artificial intelligence to help in the elimination of avoidable blindness and scale delivery of eyecare services worldwide. The consortium is working on diverse datasets of patients across geographies to come up with machine learning predictive models for vision impairment and eye diseases.

“Over the last two years we have been working with a dynamic partner ecosystem of local, regional and global organizations to bring the power of AI to Healthcare. We started with eyecare and today we are expanding into cardiac care. We have been working with our customers and partners within the healthcare sector to create AI models that can help provide doctors and health care providers insights in their treatment plans. Our partnership with Apollo Hospitals will be a major step in the journey towards providing accessible healthcare to everyone,” says Anil Bhansali, Corporate Vice President, Cloud and Enterprise, Microsoft India (R&D).

The work was jointly presented by Dr. J. Shiv Kumar, Chief of Cardiology, Apollo Hospitals; Arvind Sivaramakrishnan, Chief Information Officer, Apollo Hospitals; and Prashant Gupta, Director – Microsoft Cloud Division & Lead – AI Network for Healthcare; at HIMSS 2018 – one of the world’s largest health IT conferences.