Advancing AI in health requires access to relevant data and the tools to analyze it responsibly. T-CAIREM’s Health Data Nexus (HDN) brings these resources together, helping computer scientists, health researchers, educators, and trainees work with real-world de-identified health data.
HDN combines a data publishing platform with a cloud-based research environment. Researchers can discover datasets, complete the steps required for access, and conduct analyses within a secure environment. Data contributors can share their work with the research community while maintaining clear conditions for its use.
Our team has established a platform that supports dataset publication, researcher credentialing and cloud-based analysis. HDN provides research environments in Jupyter and RStudio, allowing users to configure computing resources for their projects. Its development and educational applications are documented in a 2025 GigaScience publication.
The platform continues to evolve. Collaborative workspaces allow authorized users to code together in the same research environment. A federated search feature, introduced in 2026, lets users discover resources on MIT's PhysioNet platform directly through HDN, broadening the range of data they can find through one interface.
HDN hosts de-identified data and maintains the processes governing access and use. These include researcher credentialing, required training, and dataset-specific data-use agreements. Additional requirements depend on the dataset and proposed activity. Research involving human participants remains subject to applicable research ethics review requirements.
A Growing Collection of Health Datasets
HDN’s hosted collection of datasets spans hospital care, medical imaging, population health, sleep laboratory data, voice recordings, clinical research summaries, and cancer biopsy images.
Examples include:
General Internal Medicine: Data from more than 22,000 patient encounters at St. Michael’s Hospital, supporting investigation of patient deterioration.
Sunnybrook Sleep Laboratory: Overnight physiological signals, sleep metrics and associated health information.
Bridge2AI-Voice: Voice recordings linked to health information for developing and evaluating machine learning methods.
Breast Cancer Digital Pathology: High-resolution slides and clinical information from 157 patients.
Immune Checkpoint Blockade: Molecular data drawn from multiple clinical trials.
Each year T-CAIREM offers Health Data Nexus Dataset Grants to support further growth by funding the creation, curation, de-identification, and documentation of datasets for research and education.
For computer scientists, HDN offers opportunities to develop and evaluate methods using real health data across multiple formats. Researchers can investigate questions involving clinical prediction, imaging, physiological signals, and other applications while working within defined data access requirements.
For health researchers, HDN provides a route to explore clinically relevant questions, collaborate with computational colleagues, and make datasets available for further investigation. Publishing data with supporting documentation helps other researchers understand its context and potential uses.
For educators, the platform supports courses and workshops that introduce learners to the complexities of real health data, including missing information, bias, and class imbalance.
T-CAIREM Health Datathons
HDN has been used for several T-CAIREM hosted Health Datathons. These events bring students, trainees, data scientists, and healthcare professionals together to investigate health questions in multidisciplinary teams.
Contribute a dataset or use HDN for teaching: To discuss dataset publication, courses, workshops, or datathons, contact our Infrastructure team: contact@healthdatanexus.ai.
