Imagine if doctors could predict who really needs an invasive colonoscopy just by looking at the bacteria in their gut. Instead of moving sensitive patient records to one giant, risky cloud, this system lets AI 'visit' different hospital databases to learn patterns without ever taking the data away. It's like a teacher visiting different classrooms to learn from students without moving the students out of their schools.
Colorectal cancer causes a 20 billion EUR annual economic burden in the EU. Current screening methods lack the specificity to accurately predict who needs expensive and invasive colonoscopies.
A prototype for local database nodes (Microb-AI-Clients) and a set of harmonized data standards for federated AI training.
If you are a medical testing laboratory dealing with low specificity in faecal immunological tests (FIT) — this project developed an AI stratification technology that aims to increase specificity by >20 percentage points using metagenomic microbiomes.
If you are a medical data software provider dealing with strict privacy regulations that hinder data integration — this project developed a federated database network that allows AI training without sensitive patient data leaving local safe harbours.
If you are a precision medicine drug developer dealing with the need for better patient stratification for therapy — this project developed computational microbiome profiles to allow for optimised personal therapy stratification in colorectal cancer.
Based on available project data, there is no specific pricing or cost information provided for the implementation of the technology.
Yes, the project aims to integrate metagenomes for over 5,000 individuals initially, with an expected total of 100,000 by 2026.
Based on available project data, specific IP and licensing details are not mentioned; however, the project involves a consortium of 8 partners including 3 SMEs.
The project uses a federated database network rather than a cloud, ensuring no sensitive patient data leaves the local databases while still allowing AI classification of phenotypes.
The project period runs from 2023-04-01 to 2028-03-31, with the goal of demonstrating results in clinical practice to help regulatory bodies adopt evidence-based guidelines.
The consortium is well-balanced for a translation project, consisting of 8 partners across 5 countries. With a 38% industry ratio (3 companies, including 3 SMEs), there is a strong focus on commercial viability and software development alongside the academic expertise of 2 universities and 2 research institutes.
Contact the University of Hamburg research office regarding the Microb-AI-ome project.
Contact us to explore licensing opportunities for federated AI in medical diagnostics.
