Anatomy of an AI Win: How Ultromics Won Over Cardiologists

November 4, 2025

Dina Bam, MPH

The buzz around artificial intelligence in healthcare is impossible to miss. From the packed halls at HLTH to the constant stream of “big announcements” in our inboxes, the promise of AI is everywhere. We hear it will revolutionize diagnostics, personalize treatment, and solve our deepest challenges.

But what happens after the announcement? What happens after the FDA clearance and the press release?

We got a front-row glimpse into two real-world implementations in a recent DiMe webinar, Earlier, Smarter, Better: Transforming Heart Failure Diagnosis with AI.

The discussion, featuring insights from cardiologists, Dr. Mike Curran (UPMC), Dr. Jordan Strom (Beth Israel Deaconess, Harvard), and Ross Upton (CEO of Ultromics), provided a perfect, high-stakes case study for the real-world challenges of AI implementation.

Cardiac amyloidosis: A problem worth solving

The conversation centered on heart failure with preserved ejection fraction (HFpEF) and one of its hard-to-diagnose culprits, cardiac amyloidosis.

Here’s the clinical dilemma:

These conditions are tough to diagnose. The symptoms are nonspecific, and as our panelists showed with a challenging visual quiz, even expert cardiologists can struggle to spot the signs of amyloidosis on a standard echocardiogram.
The stakes for missing them are catastrophic. As Dr. Curran noted, seeing a patient’s disease progression halt after a correct (and previously rare) amyloidosis diagnosis is “remarkable.”
For the first time, as Dr. Strum explained, we have transformative new therapies (like SGLT2 inhibitors) that make early diagnosis more critical than ever.

This is the perfect “AI Diagnostic Triangle,” as Ross Upton, Ultromics CEO, framed it:

A commonly misdiagnosed disease…
…that leads to a massive problem (high mortality and a $70 billion cost burden)…
…and finally has a device that can detect it early on.

AI is the key. It can see things the human eye can’t, like a “sparkly myocardium” or subtle tissue-texture changes, to flag at-risk patients years before a human could. The “why” for AI here is crystal clear.

The anatomy of an AI “win”

Drs. Strom and Curren revealed that implementing a life-saving, FDA-cleared algorithm like Ultromics into clinical practice within their health systems took 12+ months.

The real conversation wasn’t about the algorithm’s elegance and impact, and potential value to their patients. Instead, it was about the realities of AI implementation.

Pilot Fatigue & The “Signal vs. Noise” Problem: Dr. Curran described health systems as being under constant siege from “nonstop” outreaches from AI startups. How do you distinguish the one life-saving tool from the 100 others that will just be a “thorn in the side” of cardiologists?
The IT & Security Nightmare: Ross Upton detailed how a “12-month sales cycle” is the average for getting a tool like Ultromics into a health system. The real barriers are getting through IT security, navigating on-premises vs. cloud deployments, and integrating with a hospital’s PACS and EHR.
The Reimbursement Labyrinth: Dr. Strum outlined the complex maze of “who gets paid.” It’s not enough to have a CPT code. Does the payer cover it? Does the hospital know how to bill for it? And does the revenue go back to the Echolab that’s now paying for the software, or does it disappear into a general fund, making the tool look like a pure cost?
The Trust & Validation Gap: “Prove it works on my data, in my system.” Panelists agreed that even with robust FDA validation, every health system wants a local study, leading to endless one-off validations.

However, the potential wins are huge. As Dr. Curran noted, successful tech must “feel native” and not disrupt workflow. By doing the hard integration work, the Ultromics device feels like it’s just part of the cardiologist’s existing system, a massive victory for clinician adoption.

The Playbook for AI Implementation

This entire 60-minute discussion was a real-world, high-stakes case study of why DiMe convened 20+ industry stakeholders to create The Playbook Implementing AI in Healthcare.

The challenges our expert panel described aren’t unique to cardiology. They are the universal, gritty barriers that every health system leader, clinician champion, and AI developer is facing today. And the Playbook was built to be the manual for this exact fight.

When panelists talked about “signal vs. noise,” the Playbook provides a framework for assessing vendor quality, validation robustness, and readiness.
When they described the IT and security nightmare, the Playbook offers actionable guidance on governance, deployment, risk management, and integration.
When they mapped out the reimbursement labyrinth, the Playbook provides tools for building the business case, measuring ROI, and navigating payment pathways.
When they hit the wall of “pilot fatigue” and the need for local validation, the Playbook outlines strategies for building trust and evidence at scale.

The panel’s most important takeaway was a call to action for all innovators: Start with the clinical problem, not the technology.

At DiMe, we agree. The problem of heart failure is a clinical one. But the problem of getting the solution to the patient is one of implementation. AI isn’t magic; it’s a tool. And tools need a roadmap.

Though implementation of AI and other innovations into clinical practice comes with its challenges, the conversation with this incredible panel proved the “why” is worth it. Our Playbook is here to help with the “how.”

Want to learn more? Visit the AI Implementation in Healthcare Playbook resources and join our upcoming project, focused on AI-enabled Patient Care Navigation, as we continue to build the practical tools for the future of healthcare.