Medical devices & diagnostics · Cardiac imaging AI
InVision Medical Technology
Two FDA-cleared models that read heart ultrasounds already being performed, built by the lab that published the underlying papers first. Since October 1, 2026, Medicare pays hospitals extra when one of them is used.
The picture was already taken
Apical four-chamber. Parasternal long axis. Two standard views, caught in the first minutes of an ordinary heart ultrasound, on a machine the hospital already owns, by a sonographer already on shift.
Those two views are all that InVision Precision Cardiac Amyloid needs. The software sits behind the picture archive, reads the study as it lands, and writes its answer back into the same study the cardiologist is about to open, often before the scan is finished.
Nothing extra is scanned. No new hardware is bought. The bet is that something already in the picture is going unread.
“Cardiac amyloidosis often hides in plain sight on echocardiograms that are already being performed every day,” co-founder David Ouyang said in August 2026, on the day Medicare finalized a payment for finding it.
A rare disease wearing a common face
Cardiac amyloidosis is misfolded protein building up in heart muscle. Left alone it ends in heart failure and death, and its symptoms sit on top of the far more common causes of heart failure, so it is routinely missed or filed under something else. Two subtypes carry different urgency: ATTR, driven by transthyretin, and AL, driven by light chains, which needs hematology within weeks rather than cardiology within months.
That distinction used to be mostly academic. It is not any more, because targeted therapies now exist for both and work better the earlier they start. By the founders’ account, 32 million heart ultrasounds are performed in the United States each year, each taking over an hour of physician and technician time, with enough reader-to-reader variation to change what a patient is told.
The lab wrote the papers first
David Ouyang studied statistics at Rice, took his MD at UCSF, and trained in residency and cardiology fellowship at Stanford before becoming an attending cardiologist at Cedars-Sinai. He now reads echoes at Kaiser Permanente Northern California. Bryan He read computer science at Caltech and finished a Stanford CS PhD in 2022.
Between them they are authors on the papers the company is built out of: EchoNet-Dynamic, the first video-based model for beat-to-beat ejection fraction, in Nature in 2020; EchoNet-LVH in JAMA Cardiology in 2022, which classified cardiac amyloidosis at an AUC of 0.83 across 23,745 patients; and EchoNet-RCT in Nature in 2023. The dataset behind the 2022 paper, 23,212 annotated videos, was released openly, which is why outside groups can test the cleared model without asking.
“Together, we’ve worked together on this project for the last 4 years, first as a research project and now as clinical deployment of a viable product,” the two wrote when they launched out of Y Combinator in August 2022.
From paper to payment
What followed was a regulatory and reimbursement campaign more than a product campaign. An algorithm collaboration with Alexion, AstraZeneca Rare Disease in November 2023. Breakthrough Device Designation in January 2024. The ejection fraction clearance in April 2024. CPT code 0932T adopted that July, live from January 2025. The amyloid clearance in May 2025, with a competitor’s device as its predicate. HeartLab signed for Australia and New Zealand that October, a second Breakthrough designation in November. Then, in July 2026, the Medicare add-on.
What is proven, and what is still claimed
| Evidence | What the record shows | Source type |
|---|---|---|
| FDA 510(k) K232331 | Precision LVEF cleared Apr 25, 2024, product code QIH. Predicate: Caption Interpretation Automated Ejection Fraction, K210747. Retrospective multicenter validation across Philips, GE and Siemens equipment: root mean square deviation about 6.06 biplane, Dice 0.89 and 0.90. Indicated “for use in patients 22 years and older”. | Public record |
| FDA 510(k) K243866 | Precision Cardiac Amyloid cleared May 21, 2025, product code SDJ. Predicate: Ultromics EchoGo Amyloidosis, K240860. Validation used 1,221 studies from three US sites at a 1:2 case to control ratio, reaching sensitivity 0.607 and specificity 0.990, with all subgroup analyses passing. | Public record |
| What the label says | The FDA Indications for Use read: “indicated as a screening tool for adult patients aged 65 years and over undergoing cardiovascular assessment using echocardiography”. The company evidence page instead frames the clearance as an adjunctive diagnostic aid and contrasts it with screening aids. That page also gives the product code as SW0016; the FDA record says SDJ. | Differs from the record |
| The randomized trial | NCT05140642 was sponsored by Cedars-Sinai Medical Center with Ouyang as principal investigator, not by InVision. 3,495 studies, completed June 29, 2022. Nature, April 2023: substantial change by the cardiologist in 16.8% of AI-arm studies against 27.2% in the sonographer arm. The paper records no outside funding, and it tested the ejection fraction model, not the amyloid model. | Public record |
| Independent head-to-head | JACC: Advances, August 2025, led from Northwestern. 176 confirmed ATTR-CM cases against 3,192 heart failure controls. The InVision model reached AUC 0.88 against EchoGo Amyloidosis at 0.92. It scored higher on specificity and positive predictive value, lower on sensitivity, with a false negative rate of 0.43. The authors conclude: “Both deep learning echo models had a higher AUC when compared to the Mayo ATTR-CM Score”. | Independent |
| International validation | JACC: Advances, November 2025, five sites in the US and Japan. The abstract reports 574 cases, 979 controls, AUC 0.896, specificity 0.988. The company page describes the same study as 520 cases and 903 controls, all aged 65 and over, with AUC 0.893. Both are retrospective case-control. | Company figures differ |
| Medicare payment | CMS finalized an add-on payment in the FY 2027 inpatient rule of July 31, 2026: up to $2,275 per eligible stay from October 1, 2026, billed with new ICD-10-PCS code XEZZXLC, which appears in the published FY 2027 code set. CPT code 0932T has covered outpatient use since January 1, 2025. | Public record |
| Founder disclosure | The 2025 validation paper discloses that D. Ouyang “reports consulting fees and/or equity in Ultromics, InVision, EchoIQ, and Pfizer”. Ultromics builds the predicate device for InVision’s amyloid clearance and its comparator in the independent study. | Public record |
| Customers and revenue | No named US health system customer, deployment count or revenue figure was found in any public record. | Not found |
Read plainly: the science is unusually well documented for a company this small, and the regulatory scaffolding is further along than the commercial evidence. The strongest study design in the category belongs to the ejection fraction model, but that trial was run and paid for by a hospital before the company had cleared anything. The amyloid model has been tested by outsiders on their own patients, and those outsiders found it the more conservative of the two cleared devices: fewer false alarms, more missed cases. What is absent is the ordinary commercial record.
What to watch
- A prospective trial of the amyloid model. The company states the gap on its own site: “There is no prospective randomized trial of Precision Cardiac Amyloid.”
- The pathway closing behind them. Per the American College of Cardiology, the add-on alternative pathway for Breakthrough Devices is repealed beginning with 2028 applications.
- Any sign of a priced round. No SEC Form D appears under the company name.
- Precision Cirrhosis, whose American Heart Association funded trial has no registration or enrollment count on the public record yet.
- Named hospital customers, and whether the new codes are actually billed.
In their words
“Not all AI devices are the same, with many examples in the AI literature to show model performance correlates with the number of training examples.”
Bryan He, PhD, co-founder, FDA clearance release, 2024 · Company release
“Signs and symptoms of cardiac amyloidosis can resemble those of other more common diseases, often leading to delayed diagnoses or misdiagnoses and resulting in poor prognosis or incorrect treatment options.”
Seng Cheng, PhD, Alexion, AstraZeneca Rare Disease, 2025 · Partner-stated
“As a rare form of heart failure, cardiac amyloidosis is frequently missed in routine care, and AI can serve as a ‘safety-net’ for uncommon disease detection.”
David Ouyang, CEO, in HeartLab’s partnership announcement, October 2025 · Partner announcement
“Historically advanced new medical technologies have taken years to become available, particularly in places like New Zealand which are not major markets for medical device companies.”
Will Hewitt, CEO, HeartLab, October 2025 · Independent
“Every echo lab director asks the same question once the clinical considerations are settled: who pays for it.”
Andrew Reeve, Chief Commercial Officer, August 2026 · Company release
“Our AI algorithm has the highest positive predictive value (PPV) in the industry; 50% fewer false positives than the alternatives when measured head to head.”
David Ouyang, Intentional Intelligence, June 2026 · Founder blog
“In parallel, it felt like the research mission felt more at odds with the business mission.”
David Ouyang, Intentional Intelligence, April 2025 · Founder blog
Related companies
Sources
- Public recordFDA 510(k) summary and letter, K232331
- Public recordFDA 510(k) summary and letter, K243866
- Public recordClinicalTrials.gov NCT05140642
- Public recordBlinded, randomized trial of sonographer versus AI cardiac function assessment
- Public recordHigh-throughput precision phenotyping of left ventricular hypertrophy
- Public recordPerformance and potential bias of predictive models for ATTR-CM
- Public recordInternational validation of an echocardiographic AI amyloid algorithm
- Public recordFY 2027 inpatient payment final rule fact sheet
- Public recordEDGAR company and full text search
- IndependentCardiovascular device approvals lead the FDA Breakthrough Device Program
- IndependentFY 2027 ICD-10-PCS code set released
- IndependentCMS releases 2027 IPPS final rule
- IndependentHeartLab partnership announcement
- IndependentCompany profile, total raised and investors
- IndependentFunding rounds and investors
- CompanyLaunch YC: InVision: AI Diagnosis of Heart Disease
- CompanyFDA clearance and Medicare reimbursement releases
- CompanyCardiac Amyloid Evidence and Clinical Trials pages
- Founder blogSkating to where the puck is headed; My life’s work
Profile researched and written by Healthcare Discovery. Last updated September 29, 2026.
