On July 15, 2023, Sean Clifford paid $2,500 for a whole-body MRI at a Prenuvo location in New York City. Prenuvo is a California-based company that markets its scans as capable of detecting "hundreds of conditions, including most solid tumors at Stage 1, and silent killers like aneurysms." The scan was read by a radiologist operating as an independent contractor with Nexray Medical Imaging. The report came back clean: "No evidence of proximal intracranial arterial aneurysm. Normal configuration of the circle of Willis. No evidence of small vessel ischemia. No worrisome intracranial lesion is identified within the brain parenchyma." Eight months later, Clifford suffered a catastrophic ischemic stroke in Miami while on a business trip, progressing to complete occlusion of the proximal right middle cerebral artery, the same vessel and distribution covered in the July scan. He required three emergency brain surgeries and sustained permanent paralysis, vision loss, cognitive deficits, and speech problems.
Clifford's attorneys obtained copies of the original MRI images and had them reviewed by a third-party neurologist, who concluded the July scan showed "abrupt focal 60% narrowing and irregularity of the proximal right middle cerebral artery" and that the cerebral vasculature had been "incorrectly described as normal." The lawsuit, filed September 24, 2024 in the New York State Supreme Court, contends that the stenosis was visible, that intervention was available, and that a competent read of the images would have prompted referral for targeted stenting or other minimally invasive treatment that could have prevented the stroke entirely.
What's missing from the record matters just as much. Nothing documents how the radiologist actually worked the case: how long it took, which sequences they pulled up, what clinical history they had, whether Prenuvo's AI flagged anything that then got shown or buried, whether there was any rule for escalating a vascular finding they weren't sure about. About all Prenuvo has to stand on is the report itself, the one that said everything was fine. That report is now Exhibit 1 against it.
Prenuvo has tried the defenses it had. It moved to enforce an arbitration clause in its patient agreement and to apply California's malpractice damages caps instead of New York's, which has none. Courts turned down both. It also tried to keep the radiologist, Weiner, off the hook as an independent contractor rather than an employee; the court said no to that too. The case is going ahead under New York law, damages uncapped, with the radiologist still a named defendant alongside Prenuvo.
It's worth being clear about what the suit actually claims. At its core it isn't that Prenuvo's AI got the read wrong, though that's lurking in the facts. It's that the report said normal when the images allegedly didn't, and that failing to write down an obviously abnormal vessel broke from accepted radiology practice. What the AI flagged or missed, what the radiologist actually spent time on, what thinking led to the report, none of that can be answered from the record as it stands. And the empty record doesn't protect Prenuvo. It strips out the one thing that could tell a real review that reached a wrong answer from a rubber stamp that never looked at the images at all.
And the problem reaches well past Prenuvo. This is a direct-to-consumer screening product: the patient goes straight to it, around any referring physician. No ordering provider's context, no prior scans to compare against, no patient history pulled up before the read. The radiologist looks at the images and the report goes to the patient. When the pitch is catching silent killers early and the report says nothing's wrong, the patient takes that at face value. What sets the standard of care here isn't what a hospital radiology service usually does. It's what Prenuvo promised the product would do.
That gap, between what the product promised and what the record shows it delivered, is what creates the exposure. A company that sells AI-assisted detection of vascular trouble and then ships a one-page report, with no trail of how the AI and the radiologist landed on "normal," can't reconstruct its own process once someone challenges it. The evidence that would matter most here, what the AI flagged, how long the radiologist spent, which sequences they reviewed, whether any escalation path existed, isn't in the record, because nothing was set up to capture it. The no paper trail problem isn't a privacy choice in cases like this, and it isn't only a medical-AI problem. The same hole drove the FTC's action against Rite Aid's retail facial recognition deployment and sits under the class action over UnitedHealth's prior authorization AI, where a 90% appeal reversal rate became the after-the-fact proof that clinical review never really happened. The missing record didn't dodge the liability. It created it.
The case says something, too, about leaning on contract terms instead of records. Prenuvo's patient agreement had an arbitration clause the company expected to act as a backstop. It didn't survive a motion to enforce it in New York. The California damages cap it was counting on wasn't available either. Both safety nets gave way before anyone had argued the actual facts. What would have actually helped is a record made at the time, showing what the radiologist and the AI did with the July images, not a contract clause the company hoped it would never have to test.
Proof of Review captures what the Prenuvo record never had: what the radiologist looked at, what the AI showed or hid, and whether there was any way to escalate a finding nobody was sure about.