I am a pediatric cardiologist and have cared for children with heart disease for the past three decades. In addition, I have an educational background in business and finance as well as healthcare administration and global health – I gained a Masters Degree in Public Health from UCLA and taught Global Health there after I completed the program.
“Individually, we are one drop. Together, we are an ocean.” Ryunosuke Satoro, Japanese poet
Artificial intelligence in healthcare coupled with emerging technologies have high potential for healthcare dividends in the near future. We can remember the relevant technologies as:
A- Apps and applications (deep learning in protein structure prediction); B- Biomarkers (wearable, implantable, and sensor devices); C- Collective learning (federated and swarming types); D- Digital twin and thread; and E- Extended reality (augmented, virtual, and hybrid).
We covered digital twin and thread and its relevance in AI in healthcare last time in this series, and we will elucidate the relatively novel area of collective learning in this newsletter.
Until now, most AI in healthcare projects are executed mostly as local learning projects with data and computation at the same location. In addition, AI/ML have also been deployed in the cloud. Due to data privacy and regulatory concerns, however, sharing data in a central repository has been a daunting challenge and thus has severely limited collaborative efforts amongst institutions. It should be noted that this issue has improved since the onset of the COVID-19 pandemic with more institutions working closer together than before. Nevertheless, multi-institutional efforts to share data and to form large datasets have been relatively tedious and slow.
The concept of federated learning obviates the need for a centralized training dataset by having a central parameter server that will take in the weights and parameters of the models while keeping the data locally. This decoupling of the data and AI/ML can facilitate institutions’ preferences to collaborate on projects while respecting the institutions’ hesitation to share their data. This strategy ensures data privacy while accommodating models to learn collaboratively and expediently.
Swarm intelligence is the discipline of computational algorithms that is the characterized by the collective intelligence of many individual entities; this domain is inspired by nature and groups such as insects and birds. These are principles shared with complex adaptive systems previously described.
Just like swarm intelligence, swarm learning is executed in a decentralized fashion without any leader so there is no need for a central custodian. As elucidated in the Warnat-Herresthal Nature article presented a few weeks ago here at AIMed, a swarm network will have edge nodes that are used to exchange parameters for swarm learning enabled with blockchain technology. By using an application programming interface (API) to continually create an updated model, the swarm network continually “learn” and share knowledge but without sharing data.
We would like to invite you to our next exciting Clinician Series at the end of September with the focus on Primary Care (Internal medicine, precision medicine, population and global health, pediatrics, etc) with a special Healthcare Executives track.
We are also excited to welcome you to attend in-person the AIMed22 Annual Global Summit January 18th-20th, 2022, at the sublime Ritz-Carlton resort in Laguna Niguel, southern California. This summit promises to be the most enthralling yet, with Drs. Eric Topol and Daniel Kraft among the keynote speakers. We are all very much looking forward to seeing each other in person for human-to-human conversations and networking about not only exciting topics such as swarm learning but also many other topics as well!