GE HealthCare's AI Strategy: Foundation Models, Agentic AI, and a Push to 200 FDA Authorizations by 2028

How GE HealthCare is using AI in medtech - and where it's headed

GE HealthCare leads the FDA's running list of AI-enabled medical devices for the fourth year in a row. The company now counts 100 products with authorizations on that nonexhaustive list across 510(k), De Novo, and PMA pathways - and says it plans to reach 200 by 2028.

Chief AI Officer Parminder Bhatia outlines a clear path: push beyond narrow models into generative, multimodal, and agentic systems built for clinical reality. The aim is simple: less friction for clinicians, more precision for patients, and scale for health systems.

What "AI" actually means in medtech

Not all AI is equal. Imaging is a proven win: deep learning tools already help capture sharper images faster and support more confident reads. GE HealthCare's AIR Recon DL has been used in tens of millions of MR scans since 2020 to increase clarity and speed.

The step-change today comes from generative and multimodal models. Healthcare data is mostly unstructured - images, notes, audio, device signals - and traditional analytics struggle with that mix. Foundation models trained for healthcare can connect these data types to automate workflows, augment decisions, and support personalization.

What doesn't move the needle: siloed, brittle tools that can't adapt across modalities, don't explain outputs, or don't fit into the clinical workflow. In healthcare, AI has to be scalable, safe, and trustworthy - not a demo.

Vision: from add-ons to system-level change

The target problems are the ones you live with daily: staffing shortages, burnout, rising costs, and fragmented workflows. In maternal and infant care, for example, GE HealthCare's Centricity Perinatal and the Mural Clinical Intelligence Suite consolidate fetal strips, EMR data, and decision support into one view. The next step is using AI to cut documentation time and surface risk earlier so clinicians can focus on patients, not clicks.

Generative and multimodal AI can turn more of the "dark data" in healthcare into action. Agentic AI systems can work like virtual care teams - proactively coordinating, not just reacting. Underpinning all of this are Responsible AI principles: safety, fairness, and explainability.

The stakes are global. An estimated 4.5 billion people still lack access to essential health services. Smarter, more scalable tools can help close that gap.

Infrastructure that actually delivers

Healthcare-specific foundation models are central to this plan. They can handle images, clinical records, waveforms like EKG, and even genomic signals - the mix you need across screening, diagnosis, treatment, and monitoring.

GE HealthCare cites pioneering work across ultrasound (e.g., SonoSAM Track), full-body 3D MRI, and full-body X-ray. The company is also exploring multi-agentic AI with Project Health Companion, where multiple AI agents collaborate like a tumor board to synthesize complex data and propose next steps.

It's not just algorithms. Partnerships with Amazon Web Services, Nvidia, and leading academic centers (Mass General Brigham, Vanderbilt, UCSF) aim to combine imaging hardware, sensors, cloud, and compute at scale. Work with the Bill & Melinda Gates Foundation focuses on maternal and fetal health to extend impact to underserved communities.

Regulatory momentum

GE HealthCare now has 100 AI-enabled products on the FDA's public list and is targeting 200 by 2028. The list spans 510(k), De Novo, and PMA decisions and is updated as new devices are cleared or approved.

Products in practice

• Mural Clinical Intelligence Suite: Aggregates near real-time data from multiple systems and devices into a single screen for faster decisions.
• Centricity Perinatal: Supports real-time fetal strip analysis, annotations, and tight HIS/EMR integration to streamline communication and decision-making.
• AIR Recon DL: Helps radiology teams achieve sharper MR images in less time, improving throughput and diagnostic confidence.

What this means for providers and medtech teams

• Start with a high-friction workflow and define the outcome (minutes saved, accuracy gained, throughput increased).
• Insist on integration with your EMR and devices; AI that sits off to the side won't get used.
• Prioritize explainability, monitoring, and bias checks - from pilot to scale.
• Prepare for multimodal data. Imaging + notes + waveforms is where the value compounds.
• Own what differentiates you; partner for scale (cloud, GPUs, MLOps, data pipelines).
• Measure impact early and often: time to read, length of stay, alert precision, patient outcomes.

The road ahead

Foundation and agentic AI can shift care from reactive to proactive. With the right infrastructure and responsible guardrails, these systems can free up clinician time and extend high-quality care to more patients.

The work now is execution: build with clinicians, integrate deeply, validate rigorously, and scale what proves value.

FDA: AI/ML-Enabled Medical Devices list
WHO: Universal Health Coverage facts

If your team is building AI capability, explore role-based upskilling options at Complete AI Training - Courses by Job.

About the expert

Parminder Bhatia is GE HealthCare's Chief AI Officer, leading the company's strategy across foundation models, generative and multimodal AI, and responsible deployment in clinical environments.