The Aster Innovation and Research Centre, the innovation hub of the Aster DM Healthcare Group, has teamed up with Intel Corporation and AI platform provider CARPL.ai to develop and roll out an AI-powered health data platform in India.

WHAT IT DOES

The health data platform is based on federated learning, a machine learning technique that trains AI algorithms across multiple decentralised sources holding local data samples without exchanging them. 

Intel has applied OpenFL, its open source framework for training machine learning algorithms, to facilitate the adoption of federated learning. This framework has been combined with CARPL.ai’s data extract, transform, and load capabilities for end-to-end AI model training.

The Intel Software Guard Extensions have also been applied to protect workload intellectual property and secure health data. 

According to a press statement, the health data platform has been piloted using hospital data from the Kerala, Bengaluru, and Vijayawada clusters of Aster Hospitals. Over 125,000 chest x-ray images were extracted to train a CheXNet AI model using a two-site approach, which is then able to detect abnormalities in x-ray reports. 

WHY IT MATTERS

A single patient generates about 80 megabytes of imaging and EMR data each year. By 2025, the CAGR of healthcare data may reach 36%, according to a projection by RBC Capital Market. 

Although AI solutions in medical imaging have proven to be helpful in resolving pressing healthcare issues such as staff shortages, accessing silos of data across healthcare institutions, locations, and other health systems while complying with regulatory policies remains a "massive challenge," according to Aster DM. 

"Getting access to high-quality training datasets and addressing limitations in the form of regulatory frameworks and geographic boundaries are critical imperatives" in developing AI applications, said Intel India Country Head Nivruti Rai. 

By providing access to huge datasets, Aster DM's federated learning-based platform enables organisations to collaborate in developing AI-enabled health tech solutions, further boosting innovation in areas such as drug discovery, diagnosis, genomics, and predictive healthcare. It also allows clinical trials to access relevant data sets in a secure and distributed manner.

Now being offered as a service, the platform is expected to increase the accuracy of AI model training while supporting data scientists from different organisations to perform AI training without sharing raw data. With security and privacy guarantees, the platform also ensures organisational data compliance and governance.

Its recent pilot, according to Aster DM, has also shown how the platform is able to "democratise access to health data across organisational and geographical boundaries without compromising on data privacy and security aspects".

THE LARGER TREND

In recent years, the Aster DM Healthcare Group has made strides in expanding its application of AI technologies in India's healthcare landscape. A proof of this commitment is the opening of an AI lab by Aster CMI Hospital, its multispeciality hospital in Banglore. Launched in partnership with the Indian Institute of Science in March, the Aster AI lab aims to build AI healthcare tools and train healthcare professionals in AI. It will initially work on developing AI tools for neurology before expanding to other clinical specialities.

ON THE RECORD

Intel India's Rai declared that the development of the federated learning-based health data platform "marks a paradigm shift by 'getting the compute to the data' rather than 'getting the data to the compute'".

"So far, only a few such initiatives have been conducted especially in the healthcare space," claimed Dr Azad Moopen, chairman and founder of Aster DM Healthcare. He said their health data platform will "support the development of a predictive mechanism for patients, the opportunity for a second opinion on treatments, and most importantly, affirming data security and confidentiality of patients."

"There is no doubt that de-centralised data storage, and subsequent training of AI models in a federated manner, is the future, especially since lack of generalisability of AI is becoming a bigger problem," commented CARPL.ai CEO Dr Vidur Mahajan.