Share this post
The BloodCounts! Consortium, together with Flower Labs, are delighted to announce the first results of their year-long collaboration today. This collaboration brings together the largest global network of Full Blood Count (FBC) test data and research: The BloodCounts! Consortium; along with, the world’s most popular and advanced Federated Learning framework: Flower.
The BloodCounts! mission is to develop scalable diagnostic tools from routinely collected FBC test data from a growing cross-continental network of sites, with their partnership with Flower they have developed the most advanced system for secure and decentralised analysis of globally distributed FBC data. This collaboration makes possible early and accurate identification of various haematological conditions such as iron deficiency and anaemia, which affect over 1.9 billion people worldwide while still ensuring data privacy and security. The first BloodCounts! study is powered by a Flower network of hospitals in the UK, Netherlands and Gambia and has proven the feasibility of using federated learning (FL) for early haematological disease detection. This is a breakthrough in privacy-preserving international research collaborations for healthcare and enables the detection of rare conditions such as leukaemia. By the end of the year, 20 hospitals will join the network, with the rest of the consortium following in 2026, bringing the total to 108 hospitals.
BloodCounts! and Flower: A Win-Win Partnership
The BloodCounts! approach to building the world’s largest network of FBC data has been to embrace the power of FL by working with Flower Labs to deploy Flower nodes across all hospitals in the consortium network. With Flower framework, BloodCounts! can securely analyse globally distributed Full Blood Count (FBC) data without compromising patient privacy. This decentralised approach allows AI models to be trained locally within each participating institution, ensuring compliance with stringent data protection regulations and facilitating the development of robust diagnostic tools applicable across diverse populations and healthcare settings.
Flower's compatibility with various secure research environments enables seamless integration into existing healthcare infrastructures without major updates to their current systems. This adaptability accelerates the training and deployment of FL models, leading to earlier detection and treatment of haematological conditions on a global scale.
Using BloodCounts! own FL task-specific algorithm deployed with Flower, the study found improved performance for iron deficiency detection across a diverse population cohort. Newly opened hospital centres and existing ones with a high proportion of ethnic minorities achieved the most substantial gains, of over 9% in terms of balanced accuracy, while all hospital centres benefited from the federated approach. This privacy-preserving method effectively reduces healthcare disparities and minimises computational costs, as new healthcare partners can leverage pre-trained federated models without extensive retraining, promoting more equitable outcomes across globally distributed environments.
The partnership between BloodCounts! and Flower significantly advances the early detection of haematological conditions. By combining BloodCounts!'s extensive clinical and ML expertise with Flower's framework, this collaboration ensures that advancements in medical AI are achieved without compromising data privacy, paving the way for more accurate and early detection methods in healthcare.
About BloodCounts!
BloodCounts! is a global consortium of leading healthcare and research institutions working together to advance AI-driven analysis of Full Blood Count (FBC) data for early disease detection. It brings together clinicians from leading institutions such as the Cambridge University Hospitals NHS Trust (UK), NHS Blood and Transplant (UK), University College London Hospitals NHS Foundation Trust (UK), NHS Blood and Transplant (UK), Barts Health NHS Trust (UK), Amsterdam University Medical Centers (The Netherlands), Maastricht University Medical Center (The Netherlands), Zuyderland MC (The Netherlands), Health Services Authority Singapore (Singapore), Apollo Hospitals (India), The West African Centre for Cell Biology of Infectious Pathogens (Ghana) and MRC Unit The Gambia (Gambia), along with academic partners including the University of Cambridge, University College London, KU Leuven, and VU Amsterdam. BloodCounts! explores FL and machine learning to develop privacy-preserving diagnostic tools. By combining expertise from diverse healthcare systems, the consortium aims to drive innovation and improve global healthcare through data-driven research.
About Flower
Flower enables organisations and companies to train better AI models by safely leveraging distributed data. The Flower open-source framework and eco-system is the de-facto standard for both research and production around the world. It offers a unified approach to decentralized forms of learning, analytics, and evaluation; with a focus on an easy-to-use developer experience. Follow the latest updates and progress by the Flower community via Flower social media channels (e.g. https://x.com/flwrlabs,) and website.
Key contacts
BloodCounts! Consortium
- Dr. Michael Roberts: mr808@cam.ac.uk
- Dr. Nicholas Gleadall: ng384@cam.ac.uk
Flower Labs
- Dr. Javier Fernandez-Marques: javier@flower.ai
- Prof. Nicholas D. Lane: nic@flower.ai
Share this post