Clinical research environment
Standardized, scalable space to develop, train, and deploy models — with project isolation, CI/CD automation, and paths toward real clinical research settings.
Open-source platform
MAIA
Medical Artificial Intelligence Assistant
A Kubernetes-based platform for collaborative medical AI research and clinical translation — covering data preparation, model training, active learning, deployment, and evaluation in one shared environment.
Why MAIA
Bridging research innovation and clinical adoption
In medical AI, isolating individual stages of development often blocks impactful research. Fragmented workflows limit iterative improvement, validation, and — critically — translation into clinically usable solutions. A lung nodule model that never meets radiologists or hospital IT may look strong in a paper and still fail to reach PACS.
MAIA was built as a shared, open, collaborative environment where knowledge, models, and algorithms can move between researchers and institutions — and where clinical feedback is part of the loop, not an afterthought.
Capabilities
Full AI lifecycle on Kubernetes
Integrated tooling
MONAI and MONAI Bundle, 3D Slicer, OHIF Viewer, Orthanc, XNAT, MLFlow, and KubeFlow-based portals for annotation, active learning, and inference.
Federated & portable models
Support for federated learning across institutions and portable packaging via the MONet Bundle — nnU-Net inside the MONAI ecosystem for training and deployment.
Education & workshops
Hands-on environment for teaching medical AI, used by students and researchers in courses and collaborative workshops (including AIDA Tech Days).
In practice
Used in academic and clinical settings
MAIA has been a main research platform at KTH’s Division of Biomedical Imaging, supporting PhD students, Master students, postdocs, and master-level courses. A clinical deployment at the Radiology Department of Karolinska University Hospital enables closer collaboration between AI researchers and radiologists — including projects on MRI bone metastasis and brain MRI metastasis segmentation with active learning in the loop.
The platform is designed for both local and cloud-style clusters: from MicroK8s quickstarts to OpenShift in hospital infrastructure. The MAIA Toolkit provides Ansible roles and a one-command installer to stand up Core, Admin, and Dashboard layers.
Citation
Bendazzoli, S., Persson, S., Astaraki, M. et al. MAIA: a collaborative medical AI platform for integrated healthcare innovation. npj Artif. Intell. 1, 45 (2025). https://doi.org/10.1038/s44387-025-00042-6