Collective Intelligence Against Alzheimer's Disease

Alzheimer's disease is the leading cause of senile dementia, and it is extremely difficult to detect this disease in its early stages. Existing diagnostic methods are often expensive and lack sufficient accuracy, while remote healthcare facilities frequently lack access to specialized personnel and advanced equipment. BCFTL (Blockchain-enabled Multi-modal Federated Transfer Learning) – a blockchain-based framework for multimodal federated transfer learning – may offer a transformative solution.

Modern artificial intelligence can analyze vast volumes of medical data – including clinical records, imaging results, and biomarker profiles – to identify subtle patterns indicative of early Alzheimer's disease. This novel platform fully harnesses that potential. The project team comprised researchers from the University of Science and Technology of China, Hefei University of Technology, Shenyang Aerospace University (China), Korea University (South Korea), the University of Artificial Intelligence (UAE), and Togliatti State University.

“Consider a scenario where several hospitals want to train artificial intelligence (AI) to recognize early signs of Alzheimer's disease,” explains Alexey Shvetsov, one of the project's authors and project manager at TSU's Regional Engineering Project Office. “Each institution trains its local AI model using its own patient data – MRI scans, laboratory tests, clinical assessments. The data itself is not shared anywhere and remains within the hospital. This guarantees complete patient confidentiality. The hospitals then share only the learned model parameters (the AI's 'knowledge representations') with a central coordination server. There, this knowledge is combined with data from other hospitals, creating a single, smarter and more accurate overall model.” The resulting model is fed back to each hospital, making the local AI even smarter. All transactions are recorded using blockchain technology, ensuring security and transparency and making it impossible to falsify results.

“Essentially, we have created not merely a diagnostic tool, but a secure, privacy-preserving ecosystem for collaboration among medical institutions,” adds Dr. Shvetsov. “Hospitals, especially in small towns, can collectively train a powerful AI model without ever exchanging sensitive patient records. This represents a breakthrough approach for telemedicine and decentralized diagnostics.”

The platform's high efficiency has already been confirmed in experiments, achieving a diagnostic accuracy of 97% for Alzheimer's disease. The system demonstrates high throughput, processing up to 10 model updates per minute. Its primary practical application lies in extending advanced diagnostic capabilities to remote healthcare facilities worldwide. With the BCFTL platform, these institutions can now analyze local patient data while leveraging the collective intelligence of models trained across a global network – enabling reliable early detection and timely intervention.

An article describing the research by an international team of scientists was published in the world's leading scientific journal, IEEE Internet of Things Journal.

*Note: In software engineering, a framework is a ready-made set of tools that accelerates application development by providing a structured foundation for building specific types of systems.