Photo by Frederick Medina on Unsplash
- As of June 23, 2026, the FDA cleared EchoNext — developed at NewYork-Presbyterian and Columbia University — as the first AI tool to detect six types of structural heart disease from a standard ECG.
- In a head-to-head test on 3,200 ECGs, EchoNext identified structural heart problems with 77% accuracy versus 64% for cardiologists — a 13-percentage-point gap with real clinical consequences.
- Pathway Labs partnered with OpenEvidence, a clinical platform serving approximately 650,000 U.S. physicians as of January 2026, to distribute EchoNext inside existing doctor workflows rather than as a standalone hospital system.
- Pathway Labs closed an $8.5 million seed round (announced June 24, 2026) led by AlleyCorp and Breyer Capital; OpenEvidence reached a $12 billion valuation in its January 2026 Series D — both signal early-stage investor conviction in embedded clinical AI as a category.
What Happened
700,000. That is the number of matched ECG-echocardiogram pairs that trained EchoNext, Pathway Labs' new AI diagnostic tool — and it is the figure that explains why cardiologist Dr. Pierre Elias described his own creation as filling a gap that colonoscopies and mammograms never could. As of June 23, 2026, the U.S. Food and Drug Administration granted clearance for EchoNext, making it the first multi-condition AI cardiology tool to detect six distinct types of structural heart disease from a single-lead ECG (the standard electrical heart tracing that primary care doctors already order routinely).
According to Google News — drawing on reporting from TechTarget, STAT News, and a primary-source announcement from NewYork-Presbyterian — the tool was built by Dr. Elias at NewYork-Presbyterian and Columbia University, then commercialized through a spinout called Pathway Labs. On June 24, 2026, Pathway Labs announced an $8.5 million seed round led by AlleyCorp and Breyer Capital, with NewYork-Presbyterian itself participating as an investor.
The distribution strategy is where this launch diverges sharply from most medical AI rollouts. Rather than licensing exclusively to hospital systems — the traditional and notoriously slow path to clinical adoption — Pathway Labs partnered with OpenEvidence, the clinical decision platform that, as of January 2026, serves approximately 650,000 U.S. physicians and supports over 20 million clinical consultations per month. STAT News characterized the approach as a "novel step" in medical AI. TechTarget emphasized that embedding into a platform used by nearly two-thirds of American doctors removes the single biggest barrier to AI diagnostic adoption: asking physicians to change their workflow.
The Evidence Tier — What the Numbers Actually Show
Before drawing conclusions from any AI diagnostic announcement, it is worth asking what was actually measured — and this is where EchoNext's validation data stands out from the usual single-institution pilot. The head-to-head study matched the model against cardiologists on 3,200 ECGs. EchoNext identified structural heart problems with 77% accuracy; cardiologists came in at 64%. That 13-percentage-point gap is not a rounding error — at population scale, it represents thousands of patients caught earlier or missed entirely.
Chart: Structural heart disease detection accuracy — EchoNext AI vs. cardiologists on the same 3,200-ECG dataset. Source: NewYork-Presbyterian, as of June 23, 2026.
The broader validation data — provided by NewYork-Presbyterian as a primary source — is notable for its scale. EchoNext was tested across more than 20 hospitals and 500,000 patients in the U.S. and Canada, achieving a 92% performance score on standard diagnostic metrics for distinguishing patients with structural heart disease from those without. The tool can process nearly 500,000 ECGs annually, reviewing each within minutes of completion.
On June 22, 2026 — one day before FDA clearance — Nature Medicine published a case study documenting the first known heart transplant resulting from AI-detected disease: a 45-year-old patient with a rare inherited cardiac condition that EchoNext identified before any symptoms appeared. That is not a statistic; it is a proof of concept that earlier detection changes outcomes, not just spreadsheets.
In my analysis, the combination of FDA clearance, a 500,000-patient multi-hospital validation, and a published Nature Medicine case study places EchoNext in a meaningfully different evidentiary tier than most AI diagnostics currently seeking clinical adoption. The systematic review standard — large external dataset, pre-registered endpoints, independent validation sites — is largely met here. That matters for anyone evaluating health tech as part of a long-term investment portfolio, where regulatory and clinical durability are the real moat.
Dr. Elias framed the triage problem directly: "Many ECGs are abnormal. If we ordered echoes for every single abnormal ECG, we'd probably bankrupt health care." EchoNext is not designed to replace the echocardiogram (a detailed cardiac ultrasound); it is designed to identify which abnormal ECGs actually warrant one — a prioritization function that has genuine value in a resource-constrained system.
Photo by National Cancer Institute on Unsplash
Why the Distribution Model Matters as Much as the Technology
The 77% accuracy figure will attract attention. The OpenEvidence partnership is the structural story worth understanding for anyone tracking this space.
OpenEvidence raised $250 million in a Series D at a $12 billion valuation in January 2026, with total funding reaching approximately $700 million — making it among the most richly valued clinical AI platforms in history. For context, it raised at a $3.5 billion valuation earlier in the same year, then stepped up to $12 billion within months. When the American College of Cardiology announced a strategic partnership with OpenEvidence in November 2025, it set the stage for precisely this kind of embedded diagnostic deal: a validated AI tool distributed through a physician network that already trusts the delivery channel.
The analogy that fits here is how fintech companies embedded payment processing into e-commerce platforms rather than building standalone checkout systems. EchoNext, through OpenEvidence, does not ask a physician to log into a new portal, adopt new hardware, or change their ordering workflow. It surfaces inside the clinical consultation already in progress. That friction reduction is historically the difference between AI tools that achieve meaningful adoption and AI tools that gather citations without clinical impact.
For beginners thinking about financial planning and where AI health tech fits in a broader portfolio: OpenEvidence is currently private, and Pathway Labs just closed a seed round. Neither is directly investable on the stock market today. But the $8.5 million seed — anchored by a hospital system investor and led by established venture firms — reflects institutional conviction that embedded clinical AI is a durable category, not a feature waiting to be absorbed by Epic or Oracle Health.
Three Things Worth Watching
FDA clearance means EchoNext is legal to use. It does not mean Medicare, Medicaid, or commercial insurers will reimburse it automatically. Reimbursement codes for AI-assisted diagnostics remain inconsistent across the U.S. payer system. The practical adoption curve — and Pathway Labs' revenue trajectory — depends heavily on whether payer agreements follow the FDA clearance within the next 12 to 18 months. That is the milestone to watch before drawing conclusions about commercial scale.
A jump from a $3.5 billion to a $12 billion valuation within a single year (both figures from 2026 per the research data) is aggressive by any measure. That is a roughly 3.4x step-up before a potential public offering. For investors evaluating health tech exposure in their investment portfolio, the question is whether the monetization model for AI diagnostic partnerships — and how much of the $20 million-plus monthly consultation volume converts to billable AI-assisted decisions — can support that multiple at exit. EchoNext's adoption data over the next year will be an early leading indicator.
EchoNext earned the "world's first" designation for multi-condition structural heart disease detection via ECG. That lead is temporary — Google Health, AliveCor, and multiple university spin-outs have published competing cardiac AI research in adjacent domains. The durable competitive advantage is not the model, which will face capable rivals within two to three years. It is the OpenEvidence distribution network of 650,000 physicians. That embedded workflow position is the asset that is genuinely hard to replicate, and it is what makes this partnership structure worth studying regardless of whether EchoNext remains the leading tool in the category.
Frequently Asked Questions
How does AI detect heart disease from an ECG without an echocardiogram?
An ECG measures the heart's electrical activity — the wavy lines on the printout most people have seen. Structural heart problems subtly alter those electrical patterns in ways that are statistically learnable but visually difficult to catch by eye. EchoNext was trained on over 700,000 paired ECG-echocardiogram records, so it learned which subtle ECG signals correlate with confirmed structural abnormalities later verified by echocardiogram. It then applies those learned associations to new ECGs in real time, flagging cases that likely warrant full echocardiographic follow-up. As of June 23, 2026, the tool processes nearly 500,000 ECGs annually, reviewing each within minutes of completion.
Is AI accurate enough to trust for heart disease detection in a real clinical setting?
For EchoNext specifically, the validation evidence is more robust than most AI diagnostics currently in clinical use. It was tested across more than 20 hospitals and 500,000 patients in the U.S. and Canada — closer in scale to a Phase 3 clinical trial than a typical academic pilot. It achieved a 92% performance score on standard detection metrics and outperformed cardiologists 77% to 64% on a 3,200-ECG head-to-head test, per NewYork-Presbyterian's primary source data. No diagnostic tool is 100% accurate, and EchoNext is designed as a triage assistant — flagging high-risk cases for follow-up — rather than as a replacement for specialist judgment or confirmatory imaging. Talk to your physician about what any ECG result means for your specific situation.
Can AI replace cardiologists in diagnosing heart conditions in the near future?
The current evidence — and Dr. Elias' own framing of EchoNext — positions AI as a prioritization layer rather than a replacement. The bottleneck the tool addresses is not "cardiologists make errors"; it is that most patients with structural heart disease never see a cardiologist until they are already symptomatic, because there has been no cost-effective screening bridge between a routine ECG and a full echocardiogram. AI handles the volume triage; cardiologists handle confirmation and treatment decisions. As of June 25, 2026, published evidence supports augmentation at population scale — but the clinical workflow, the liability framework, and the payer structure all still assume a physician in the loop for final diagnostic decisions.
Disclaimer: This article is for informational and educational purposes only and does not constitute financial or medical advice. All figures and claims are based on editorial commentary on publicly reported news and research. Readers should consult a licensed financial advisor before making investment decisions and a qualified healthcare provider for any medical concerns. Research based on publicly available sources current as of June 25, 2026.