Periodic Reporting for period 1 - ODELIA (Open Consortium for Decentralized Medical Artificial Intelligence)
Periodo di rendicontazione: 2023-01-01 al 2024-06-30
In the rapidly advancing field of healthcare, innovation is the driving force behind improving patient outcomes and expanding medical knowledge. The Open Consortium for Decentralized Medical Artificial Intelligence (ODELIA) is an EU-funded project poised to significantly advance healthcare through the use of Swarm Learning (SL). ODELIA is dedicated to harnessing the power of SL to develop artificial intelligence (AI) solutions for breast cancer detection in MRI screenings while prioritizing data privacy and fostering collaboration.
In an age where data privacy is paramount, ODELIA offers a groundbreaking approach to AI development. By uniting institutions in a pan-European SL Network, ODELIA facilitates secure and collaborative AI development without the need to share sensitive patient data. This ensures that patient privacy remains intact—a crucial ethical consideration in healthcare AI.
ODELIA’s primary objective is to build the first open-source software framework for SL, providing an assembly line for the streamlined development of AI solutions. Similar to federated learning, SL is a decentralized approach that enables machine learning models to learn on distributed data without sharing raw patient data. SL however does not require a central hub or server, which is typically required for federated learning that involves aggregating federated models to update a global model.
The project partners collaborate to develop a clinically useful AI algorithm for the detection of breast cancer in MRI, using a distributed database. Breast cancer is a leading cause of mortality among women, and early diagnosis is pivotal for improving survival rates. Using SL, ODELIA aims to enhance diagnostic performance, accelerate AI development, and create robust, generalizable solutions for better healthcare outcomes.
The ODELIA project’s success is expected to have far-reaching impact. It will not only deliver a useful medical application for the detection of breast cancer but also serve as a model for similar initiatives in other medical fields. By fostering secure, collaborative AI development and protecting patient data, ODELIA paves the way for a new era of healthcare innovation, thereby promoting and fostering transparency and trust in AI solutions in healthcare.
In an age where data privacy is paramount, ODELIA offers a groundbreaking approach to AI development. By uniting institutions in a pan-European SL Network, ODELIA facilitates secure and collaborative AI development without the need to share sensitive patient data. This ensures that patient privacy remains intact—a crucial ethical consideration in healthcare AI.
ODELIA’s primary objective is to build the first open-source software framework for SL, providing an assembly line for the streamlined development of AI solutions. Similar to federated learning, SL is a decentralized approach that enables machine learning models to learn on distributed data without sharing raw patient data. SL however does not require a central hub or server, which is typically required for federated learning that involves aggregating federated models to update a global model.
The project partners collaborate to develop a clinically useful AI algorithm for the detection of breast cancer in MRI, using a distributed database. Breast cancer is a leading cause of mortality among women, and early diagnosis is pivotal for improving survival rates. Using SL, ODELIA aims to enhance diagnostic performance, accelerate AI development, and create robust, generalizable solutions for better healthcare outcomes.
The ODELIA project’s success is expected to have far-reaching impact. It will not only deliver a useful medical application for the detection of breast cancer but also serve as a model for similar initiatives in other medical fields. By fostering secure, collaborative AI development and protecting patient data, ODELIA paves the way for a new era of healthcare innovation, thereby promoting and fostering transparency and trust in AI solutions in healthcare.
During the first reporting period the project embarked on establishing a robust decentralised platform, laying the foundation for collaborative AI model training across multiple institutions without centralised data sharing. A comprehensive manual for implementing SL on-site was compiled, encompassing hardware and software requirements, preprocessing workflows, and seamless integration into existing workflows.
The partners demonstrated the practical application of SL in breast cancer detection using MRI data, marking a significant milestone. The team released a minimal viable product, showcasing decentralised model training capabilities and its efficacy in utilising publicly available data.
A technical documentation website for ODELIA’s open-source SL implementation was set up. This site serves as a valuable resource for those looking to leverage ODELIA’s implementation for diverse use cases, promoting collaborative innovation. The documentation is openly accessible at odelia-ai.github.io.
Furthermore, strides have been made in enhancing the user experience with the development of a user-friendly front-end. This facilitates seamless data gathering for new AI model development while bridging the crucial gap to clinical usability.
First experiments of training a joint model on distributed data from three partners, and publicly available datasets have been performed.
The partners demonstrated the practical application of SL in breast cancer detection using MRI data, marking a significant milestone. The team released a minimal viable product, showcasing decentralised model training capabilities and its efficacy in utilising publicly available data.
A technical documentation website for ODELIA’s open-source SL implementation was set up. This site serves as a valuable resource for those looking to leverage ODELIA’s implementation for diverse use cases, promoting collaborative innovation. The documentation is openly accessible at odelia-ai.github.io.
Furthermore, strides have been made in enhancing the user experience with the development of a user-friendly front-end. This facilitates seamless data gathering for new AI model development while bridging the crucial gap to clinical usability.
First experiments of training a joint model on distributed data from three partners, and publicly available datasets have been performed.
ODELIA has successfully demonstrated the potential of SL in improving breast cancer detection models. By leveraging distributed data from multiple hospitals, we have developed models that outperform those trained on single-institution datasets. This approach not only enhances model performance but also addresses critical issues of data privacy and security in healthcare AI development. Our results suggest that SL could be a key technology in advancing the field of medical imaging AI, particularly in areas characterized by non-standardized imaging protocols and diverse patient populations. Moving forward, we aim to improve these first models and to expand our research to include additional imaging modalities and explore the application of swarm learning in other areas of healthcare, with the ultimate goal of improving patient outcomes through more accurate and generalizable AI-assisted diagnostic tools.