Expanding Healthcare AI Models to New Modalities and Capabilities

At HIMSS 2025 Microsoft introduced several new breakthrough developments in multimodal healthcare AI to ensure that healthcare developers and researchers have the tools and expertise necessary to innovate the use of AI in building the healthcare AI applications of the future. 

Fine-tuning capability for MedImageInsight

To further help health AI model developers fine-tune and customize our most popular state of the art multimodal foundation model, we’ve introduced a fine-tuning training capability for MedImageInsight in Azure Machine Learning. This enhancement lowers barriers to building and testing robust imaging classifiers and embedding models tailored to specific settings.

This will enable building on the performance for 14 medical imaging modalities. Even with smaller, highly specialized datasets, developers can now adapt our pre-existing models to explore the capability to detect nuanced patterns in clinical imaging—such as subtle tumor markers or rare pathologies. By streamlining the process from model selection to testing and validation, Azure AI Foundry equips innovators with a nimble framework that accelerates discovery and shortens development cycles.

Use Case: MedImageInsight in Action

MedImageInsight is already showing the potential for impact in healthcare research. As Alan McMillan, Professor of Radiology at the University of Wisconsin School of Medicine and Public Health, shared:

              “We optimized the sensitivity and specificity of Microsoft’s MedImageInsight foundation model to accurately identify a significant portion of normal chest X-rays. This could allow radiologists to focus more on analyzing suspicious cases. In our evaluation, we found that at a threshold achieving 99% sensitivity for detecting abnormal chest X-rays, the model could reduce radiologists’ workload on normal cases by 42%.”

This kind of workflow optimization would be a crucial step toward reducing radiologist burnout and ensuring that specialists can dedicate more time to the most clinically significant cases.

Expanding the Health and Life Science AI Model Catalog

Within our model catalog in Azure AI Foundry, we’re introducing several new and updated multimodal medical foundation models to our industry-leading ecosystem:

• Rad-Dino – A radiology-focused Chest X-Ray foundation model with over 1M downloads on Hugging Face and published in Nature Machine Intelligence.
• TamGen – A protein design model that employs a chemical language model to generate target-specific molecules, accelerating drug discovery and published in Nature.
• BioEmu-1 – A protein design model that generates diverse structural conformations of proteins, enhancing understanding of protein dynamics and function.
• Hist-ai – Cutting-edge pathology models
• ECG-FM – A foundation model for electrocardiogram (ECG) analysis.
• MedImageParse 3D – An adaptation of Microsoft’s MedImageParse model (now called MedImageParse 2D) for 3D medical imaging.

Continuing our commitment to open-source AI, many of these models will also be available on Hugging Face, broadening accessibility and fostering a global community of researchers and innovators.

The Microsoft healthcare AI models are intended for research and model development exploration. The models are not designed or intended to be deployed in clinical settings as-is nor for use in the diagnosis or treatment of any health or medical condition, and the individual models’ performances for such purposes have not been established. You bear sole responsibility and liability for any use of the healthcare AI models, including verification of outputs and incorporation into any product or service intended for a medical purpose or to inform clinical decision-making, compliance with applicable healthcare laws and regulations, and obtaining any necessary clearances or approvals.

Partnering with Stanford: The First Public Healthcare AI Model Leaderboard

Alongside these core technology offerings, we are partnering with Stanford University to support their launch of the first public Healthcare AI Model Leaderboard building upon real-world clinical data and use cases. This initiative shines a spotlight on these innovative models, comparing performance across a diverse range of benchmarks that go beyond multiple-choice clinical testing data and bring much-needed transparency to the healthcare AI space, allowing researchers to rapidly identify which models align best with their particular use cases. It also fosters a spirit of open, data-driven competition that we believe will push the industry toward increasingly accurate and reliable solutions.

HIMMS 2025

We're thrilled by the enthusiastic response we received when showcasing these groundbreaking advancements at HIMSS 2025. It was inspiring to share detailed insights and see firsthand the excitement around innovations like the fine-tuning capabilities for MedImageInsight, the expanded catalog featuring new models such as Rad-Dino, TamGen, BioEmu-1, Hist-ai, ECG-FM, and MedImageParse 3D, as well as our pioneering partnership with Stanford University on the first public Healthcare AI Model Leaderboard.

Our unified, secure, and high-performing platform, designed to power the next generation of healthcare, generated significant excitement and valuable conversations, reinforcing our commitment to accelerating innovation, improving patient outcomes, and driving healthcare forward.

None of this would be possible without our incredible network of collaborators—research institutions, clinical partners, technology innovators, and dedicated healthcare practitioners. Your ongoing partnership continually fuels and challenges our progress.

Thank you to everyone who joined us at HIMSS 2025—here's to a future shaped together, powered by AI-driven solutions that meaningfully reduce administrative burdens, enhance diagnostic precision, and ultimately save more lives.