The three hospitals needed a faster way to review chest X-rays during COVID surges because expert interpretation was time-intensive and specialist capacity was limited.
Reported impact
The consortium delivered a privacy-preserving diagnostic proof of concept that improved model performance by 24% over local baselines and created a practical foundation for helping additional hospitals handle diagnostic surges with less specialized expertise.
Hospital Ramón y Cajal, Hospital 12 de Octubre, and Hospital San Pablo worked with Capgemini and a wider technology consortium to build a privacy-preserving AI proof of concept for COVID diagnosis from chest X-rays. Each hospital kept patient data locally while the hospitals and Capgemini developed a federated learning platform that let experts train computer vision models on scarce data sets and combine the learning into a single aggregate diagnosis model. Cisco and Intel supplied the computing environment for experimentation, and Intel Xeon Scalable processors with Software Guard Extensions protected encrypted aggregation tasks so identifiable patient data did not leave the hospitals. The resulting model screened symptomatic patients for COVID status, outperformed the best local models by 24%, and showed a path to helping hospitals with less X-ray expertise handle surges without compromising privacy.
Each hospital keeps chest X-ray data and local expertise on site, local training feeds a federated learning platform, encrypted aggregation on Intel SGX combines learning across the three hospitals, and the resulting computer vision model screens symptomatic patients for COVID without transferring identifiable patient information.
"Compared to the best available local models, the aggregate model achieved a 24% increase in diagnostic performance while leveraging common data."
"In the end, we made sure that patient data never lever left the hospitals."
"This included a federated learning platform, developed through a collaboration between the hospitals and Capgemini, that would enable experts to train computer vision models with scarce data sets."
"Together, the consortium members set about creating a proof-of-concept that would use artificial intelligence and computer vision to screen X-rays and determine a symptomatic patient’s COVID status."
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
The official Madrid regional government page states the hospital is 'un Centro de titularidad pública, dependiente de la Consejería de Sanidad de la Comunidad de Madrid e integrado en el Servicio Madrileño de Salud -SERMAS-.'
The official Madrid regional government page describes Hospital Universitario 12 de Octubre as one of the hospitals of the Servicio Madrileño de Salud and notes its 'más de 6.000 profesionales.'
The official Sant Pau page says the hospital has 'más de 3.000 profesionales' and is governed by a foundation board where 'están representados la Generalitat de Cataluña, el Ayuntamiento de Barcelona y el Arzobispado de Barcelona.'
Related
Capgemini Sample Mar 17, 2025 Fresenius Digital Technology Fresenius Digital Technology needed a scalable global IT support channel that improved user access and ticket resolution without adding a separate external technology stack. More than 130,000 FreDi interactions since the initial go-live in September 2023 FreDi · Microsoft Teams · ServiceNow Capgemini Sample Jul 13, 2023 Boehringer Ingelheim Boehringer Ingelheim had millions of unstructured text documents in inconsistent formats and quality states, making it hard for teams across the business to analyze the data, surface insights, and respond quickly to… 40% reduction in response time for complex medical inquiries Custom NLP-driven analytics platform · Machine Learning Operations approach · Natural Language Processing (NLP) algorithms 
Google Cloud Sample Archived · Oct 9, 2026 Angelus Founded by a dentist whose first product was based on a patent, Angelus competes on differentiation, developing bioceramic products and fiberglass post lines that require clinical proof, intellectual property… 50,000 pages of proprietary data now powering AI-generated documents Gemini 