🧠Galeon AI®

AI is revolutionizing healthcare

Instantly detecting diseases
Choosing the most suitable treatment for each patient
Learning from millions of patient records
Discovering new drugs automating caregivers' administrative tasks

The applications of AI in healthcare have revolutionary potential, and we are only at the beginning! To create AI, innovators and engineers need large quantities of structured data.

Dr. Nozman details these innovations in this video:

Galeon structures and stores a phenomenal amount of medical data. These pieces of information remain the property of the patient.

At any time with Galeon, a patient who has given consent for the use of their data can retract it.

Galeon also respects and protects hospitals' sovereignty over their data by building a decentralized collaborative network for AI training called Blockchain Swarm Learning®.

Blockchain Swarm Learning®

Galeon has developed proprietary technology for transparent and secure AI training. The Galeon blockchain serves to connect hospitals to decentralize medical research among multiple hospitals distributed worldwide.

Data is the fuel for training Artificial Intelligences: it is a rare, valuable, and sensitive resource.

To access medical data and train AIs, the GAFAM and AI startups propose several learning architectures. We can distinguish 4 main types of AI training: localized, centralized, federated, and swarm-trained by blockchain (Blockchain Swarm Learning®). Each type of AI training has its pros and cons.

Localized Learning

Each hospital trains its AIs on its own data.

Localized learning may be enough for AI applications requiring minimal data and resources. This may be suitable for simple applications developed within any hospital. However, limitations are quickly reached for the development of data-driven personalized treatments.

Centralized Learning

Each hospital entrusts its data to a third-party platform that trains AIs.

Centralized learning allows for pooling data resources at an AI expert company, but sovereignty over the data and research results is not guaranteed for hospitals.

This model is not accepted by hospitals and governments.

Federated Learning

The data remains the property of each hospital. However, because the data is heterogeneous, a single access point is still needed to clean it. The AIs are then owned by a third-party platform.

Federated learning is a variant of centralized learning, where this time the data is supposed to remain the property of the hospitals.

However, the heterogeneity of data at the care level is a major obstacle to training AIs on large databases. The need for post-processing reduces the weight of hospitals in the healthcare data value chain.

Limitations of localized, centralized, and federated learning

The healthcare applications of AI are becoming more precise every day.

They increasingly require data in both quantity and quality (structured). Localized, centralized, and federated learning approaches are all limited at a certain scale by challenges related to skills resources, infrastructure, data heterogeneity, or regulatory barriers, and thankfully, for privacy reasons.

It is for the development of the most precise applications that Galeon is building Blockchain Swarm Learning®, a way to train AIs on data distributed across multiple continents without compromising data privacy.

Blockchain Swarm Learning®

The training takes place on a decentralized set of data across multiple hospitals. The data remains on the hospitals' servers and is not put on a blockchain.

The Galeon blockchain between hospitals is used to trace the training of AIs on the data they provide and to distribute the value created proportionally to the data used.

Blockchain Swarm Learning® (BSL®) protects patient data confidentiality.

Why Blockchain Swarm Learning® (BSL®) ?

Galeon has chosen an approach faithful to the fundamental principles of Bitcoin:

"Don't trust, verify"

The Galeon blockchain between hospitals, where hospitals are both users and validators, ensures complete respect for patient privacy and data integrity.

Data is hosted locally by each hospital, with only the artificial intelligence algorithms 'moving' via the blockchain to be trained in a decentralized manner.

Thanks to the medical blockchain, all actions are traced during AI training.

The value created by AI must be fairly shared among patients, innovators, investors, and caregivers for a sustainable system. Each use of health data by BSL® results in a transaction that will be distributed as follows:

50% to the Galeon DAO fund, consisting of voluntary patients and Galeon pioneers
40% to the hospitals, where the data is captured and structured
5% for buyback and burn
5% for Galeon

This distribution ensures balance and sustainability of the system by sufficiently incentivizing all stakeholders according to their contribution.

For example

A startup funded by venture capital (VC) raises 20,000,000 USD to train its models on healthcare data. It spends 10,000,000 USD for access to data on the Galeon platform.