Heartvue.ai has received U.S. Food and Drug Administration 510(k) clearance for Heartvue.Proton, a machine-learning-powered cardiac MRI platform designed to automate quantitative measurements that can consume substantial physician and technologist time during conventional analysis. The software can assist with 2D linear measurements, ventricular volumes and ejection fraction, and blood-flow quantification while keeping final review and interpretation under the control of qualified clinicians. Heartvue plans to begin commercial rollout to U.S. hospitals and imaging centers following the clearance.
The FDA database records Heartvue.Proton under 510(k) K260811 as automated radiological image-processing software, with a decision date of August 26, 2026. The agency also authorized a Predetermined Change Control Plan, or PCCP, allowing specified future machine-learning modifications to be implemented within preauthorized boundaries without requiring a completely new marketing submission for every covered update.
What does Heartvue.Proton actually automate?
Cardiac MRI can provide highly detailed measurements of heart structure, ventricular function and blood flow, but generating those quantitative results traditionally requires substantial manual contouring and measurement. Heartvue says analysis of one examination can take up to an hour or longer in some workflows, limiting how easily quantitative MRI can be scaled across busy cardiology and radiology departments.
Heartvue.Proton automates measurements across three main categories. It can support 2D linear dimensions, ventricular volumes and ejection-fraction calculation, and quantitative assessment of blood flow. The software is designed to integrate with hospital picture-archiving systems and electronic health records so that clinicians can use it within existing reading workflows rather than exporting cases into a completely separate environment.
Automation does not mean the algorithm becomes the interpreting physician. The platform is intended to assist qualified professionals, who remain responsible for reviewing the measurements, correcting them when appropriate and incorporating them into the complete clinical assessment.
Why is ejection fraction such an important cardiac measurement?
Ejection fraction describes the proportion of blood pumped out of a ventricle with each contraction and is one of the most widely used measures of ventricular systolic function. Reduced left ventricular ejection fraction can influence the diagnosis and management of heart failure, cardiomyopathy and numerous other cardiac conditions.
Cardiac MRI is particularly valuable because it can provide highly reproducible ventricular-volume measurements while also supplying tissue characterization that echocardiography cannot always match. The limitation is that manually tracing ventricular borders across multiple image slices can be time-consuming, creating an ideal task for automation if the software remains reliable across different patient anatomies and image quality.
Blood-flow quantification presents a similar opportunity because cardiac MRI can measure velocity and flow across vessels and valves. Automating those calculations can potentially make advanced quantitative information available more consistently without requiring every imaging center to maintain the same level of manual post-processing expertise.
What is unusual about the Predetermined Change Control Plan?
Traditional medical-device regulation can create a challenge for machine-learning systems because an algorithm may improve as developers obtain additional clinical data. If every meaningful update required an entirely new regulatory submission, software innovation could move more slowly than the underlying technology.
A PCCP allows a manufacturer to tell the FDA in advance which types of modifications it may make, how those changes will be developed and validated, and what controls will prevent them from creating new unacceptable risks. Once the agency authorizes that plan, specified updates can proceed within those boundaries without requiring another 510(k) submission each time.
The authorization therefore does not give Heartvue unrestricted permission to change its model however it wants. Material modifications outside the approved plan can still require additional regulatory review. The importance is that the regulatory lifecycle begins accommodating controlled algorithm evolution rather than treating software as permanently frozen on the day of clearance.
How quickly did Heartvue reach FDA clearance?
Heartvue and regulatory-development partner Innolitics say the 510(k) was cleared approximately five and a half months after submission. The FDA database lists the application as received on March 12, 2026 and substantially equivalent on August 26.
It was Heartvue.ai’s first FDA submission, making the combination of initial clearance and an authorized PCCP particularly notable for a relatively young cardiac-imaging company. Innolitics supported the regulatory strategy, clinical-validation design, cybersecurity documentation, machine-learning documentation and PCCP framework.
Could AI make cardiac MRI more widely available?
The scanner itself is only one bottleneck. Cardiac MRI also requires specialized acquisition protocols, technologists familiar with cardiac imaging and clinicians capable of interpreting large datasets. Time spent on manual post-processing can further limit throughput, particularly outside major academic centers.
Automation could make quantitative MRI easier to scale by reducing repetitive measurement work and giving clinicians more time for interpretation. If analysis time falls substantially without sacrificing accuracy, imaging centers may be able to process more studies or offer advanced cardiac MRI in settings where expert staff are limited.
That does not mean software alone solves access. MRI scanner availability, examination duration, patient contraindications and reimbursement still determine how many patients can undergo the test.
How should clinicians think about errors from an automated MRI system?
Every automated measurement system can make mistakes, particularly when anatomy is abnormal, image quality is poor or the case falls outside patterns well represented in training data. That is why the Heartvue system retains physician review rather than presenting its output as autonomous diagnosis.
The long-term test will be how frequently clinicians need to edit automated measurements and whether performance remains reliable across scanners, field strengths, institutions and diverse cardiac diseases. Commercial deployment should generate a much larger real-world performance base than was available during initial regulatory validation.
The PCCP becomes relevant here because improvements based on postmarket experience can potentially be introduced through a predefined regulated pathway rather than leaving the original algorithm unchanged indefinitely.
Why does this matter beyond one cardiac MRI product?
Medical imaging AI has spent years proving that algorithms can segment anatomy and calculate measurements. The next commercial challenge is integrating those systems deeply enough into routine workflow that clinicians actually save meaningful time.
Heartvue.Proton is entering the market with two characteristics aimed directly at that problem: PACS/EHR integration and regulatory permission for defined future model improvements. If those elements work as intended, the platform becomes less of a standalone AI demonstration and more of an evolving component inside the clinical imaging stack.
Its success will ultimately be measured less by whether machine learning can draw ventricular borders and more by whether hospitals find that the automation reduces workload, expands access to quantitative cardiac MRI and remains trustworthy enough that physicians allow it to become routine.
