Aster Innovation and Research Centre, the innovation hub of Aster DM Healthcare, has partnered with Intel Corporation and CARPL to announce a state-of-the-art “secure federated learning platform”. This collaboration will enable the development of AI-based health technology solutions where data can safely reside where it is generated. This collaboration will drive innovation in areas such as drug discovery, diagnostics, genomics and predictive healthcare. It will also allow clinical trials to access relevant datasets in a secure and distributed way.
A single patient generates nearly 80 MB of data per year in imaging and EMR data; according to 2017 estimates, RBC Capital Market predicts that “by 2025, the compound annual growth rate of healthcare data will reach 36%. Genomic data alone is expected to be 2 to 40 exabytes by 2025, dwarfing the amount of data acquired by all other technology platforms.
AI-based solutions in areas such as medical imaging are helping address pressing healthcare challenges such as staffing shortages and an aging population. However, accessing relevant data silos distributed across hospitals, geographies, and other healthcare systems while complying with regulatory policies is a daunting challenge.
Commenting on this one-of-a-kind collaboration, Dr. Azad Moopen, Founder, President and CEO of Aster DM Healthcare, said, “Aster’s Innovation and Research Center is pleased to partner with giants of technology like CARPL and INTEL to bring highly progressive healthcare solutions through digital advancements and artificial intelligence. The Secure Federated Learning initiative will help analyze data and support the development of a predictive mechanism for patients, the opportunity for a second opinion on treatments, and most importantly, affirmation of data security and patient confidentiality. So far, only a few such initiatives have been carried out, especially in the area of health. This collaborative platform with world leaders will open the doors to many players in the sector to participate in the development of accessible health solutions.
Nivruti Rai, Country Head, Intel India & Vice President, Intel Foundry Services, said, “AI applications are poised to revolutionize healthcare through rapid and effective screening, diagnosis and treatment of diseases. . Access to high-quality training datasets and resolution of limitations in the form of regulatory frameworks and geographic boundaries are key imperatives. I am happy to report that Aster and Intel have collaborated to address these challenges and deployed a one-of-a-kind secure federated learning platform in India. It offers a concrete solution by addressing key aspects such as security, trust and privacy for optimal use of data. This solution will be offered as a service to be used by both AI researchers and data custodians in their quest to advance AI innovation and have a significant impact on healthcare. This marks a paradigm shift in bringing computation to data rather than bringing data to computation. Our common intention is to make this platform available to the health ecosystem to solve some of the large-scale health problems and enable quality, affordable and large-scale health care.
Dr. Vidur Mahajan, President and CEO of CARPL.ai, said: “There is no doubt that decentralized data storage and subsequent training of AI models in a federated manner is the future, especially that the lack of generalizability of AI becomes a bigger problem.We are pleased to partner with brands that are deans of their respective fields – Aster in healthcare and Intel in compute – to enable data extraction, anonymization, annotation and delivery to AI models via CARPL Our mission is to take AI from the test bench to the clinic, and this is another example of the same thing.
How it works
Federated learning (FL) is a method of training AI algorithms with data stored in multiple decentralized sources without moving that data. To facilitate the adoption of federated learning, Intel led the development of the OpenFLopen source framework for training machine learning algorithms, which provide a solution to “data silos” by leveraging the security technology of ‘Intel.
Intel® Software Guard Extensions (Intel® SGX) provides hardware-based memory protection by isolating specific application code and data in memory. This secure FL solution enables workload intellectual property (IP) protection and secures healthcare data with its custodians. OpenFL was combined with CARPL’s rich data extraction, transformation and loading (ETL) capabilities for end-to-end AI model training.
The demonstration of the capability of this platform was carried out using hospital data from the Kerala, Bengaluru and Vijayawada clusters of Aster Hospital. Over 125,000 chest X-ray images, including 18,573 images selected from over 30,000 unique patient data from Bengaluru, were used to train a CheXNet AI model using a two-node/site approach – Bengaluru and Vijayawada – Federated learning to detect anomalies in the X-ray report. The 18,573 unique frames, in addition, provided a 3% improvement in accuracy due to real-world data that was otherwise unavailable for training the AI model.
Health Ecosystem Benefits
- Enables data scientists from multiple organizations to conduct AI training without sharing raw data
- Enables healthcare providers and other ecosystem partners to access larger datasets used to develop AI models used in preventive and predictive medicine
- Ensures compliance and governance of organizational data, as data is not shared due to security and privacy safeguards
- Increases accuracy of AI model training through access to larger datasets.
The success of this pilot project has demonstrated commitment to the next level, which is to democratize access to health data across organizational and geographic boundaries without compromising privacy and data security aspects.