Advancing Federated Learning towards Real World Medical Image Analysis

The Flower Health & Life Sciences Day at Jesus College, Cambridge brought together leading researchers and practitioners to discuss the growing role of federated learning and responsible AI in healthcare and life sciences. Talks from the Flower team highlighted new facilities and opportunities for large-scale collaboration across health and pharma, including advances such as Flower Supergrid and the newly introduced Confidential Compute feature for secure AI training.

Our postdoctoral researcher, Pramit Saha spoke about two key aspects of our Federated Learning works done in the Noble Lab: 1️⃣ Our 𝗶𝗻𝘁𝗲𝗿𝗻𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗰𝗼𝗹𝗹𝗮𝗯𝗼𝗿𝗮𝘁𝗶𝗼𝗻 between hospitals and research organisations across #India, the #UK, and #Australia, developing 𝗳𝗲𝘁𝗮𝗹 𝘂𝗹𝘁𝗿𝗮𝘀𝗼𝘂𝗻𝗱 𝘃𝗶𝗱𝗲𝗼 𝗺𝗼𝗱𝗲𝗹𝘀 via Federated Learning, and 2️⃣ Our efforts to enable real-world FL collaboration by minimising human intervention and incorporating 𝗮𝗴𝗲𝗻𝘁𝗶𝗰 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸𝘀 𝗶𝗻𝘁𝗼 𝘁𝗵𝗲 𝗙𝗟 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄. The talk was very well received — the audience showed great interest and asked engaging questions about our approach and ongoing projects. 

Invited speakers from BloodCounts!, PharosAI, King’s College London, Eye2Gene, and the University of Cambridge presented diverse applications of federated learning — from global blood analysis and cancer research to surgical AI, ophthalmic diagnostics, and clinical data studies. It was inspiring to learn from others’ efforts — from federated learning in surgical data science, to multi-country ophthalmology and COVID studies, and even multimodal foundation models that integrate imaging and blood data. The diversity of use cases shows how rapidly the FL community is evolving.

If you’d like to watch the full talk, the recording is now available on Flower’s YouTube channel.