Half of all heart attacks occur in patients previously classified as low-risk.
Aitheroscope analyzes eye fundus images to detect atherosclerosis before symptoms appear, with validation showing 5% false positives and 3% false negatives versus clinical standards. The Spanish startup has deployed across 18 clinical centers nationally and signed distribution agreements for Latin America, expanding access to early cardiovascular screening.
- Aitheroscope achieved 5% false positives and 3% false negatives in validation against carotid ultrasound in over 1,000 patients
- Deployed across 18 clinical centers in Spain; distribution agreements signed for Mexico and Chile
- Team of 10 people: cardiologist, computer vision specialist, AI experts, and business leaders
- Approximately 50% of heart attacks occur in patients previously classified as low-risk, particularly women
Horus ML's Aitheroscope uses AI to detect subclinical atherosclerosis from retinal images in primary care, achieving 95% accuracy and addressing widespread underdiagnosis of cardiovascular disease.
Jesús Prada Alonso spent a decade working in healthcare, watching cardiovascular disease move through his own family. He knew artificial intelligence could help—but only if it solved a real problem, not just optimized a metric. When Carolina Espejo, one of his cofounders and an interventional cardiologist, brought him a specific crisis, the shape of Aitheroscope became clear.
Espejo had lived the problem firsthand. Subclinical atherosclerosis—the silent buildup of plaque in arteries before any symptom appears—was being missed at scale. Roughly half of all heart attacks occur in patients who had been classified as low-risk. The underdiagnosis was particularly severe in women. There was a gap between what doctors could detect and what was actually happening inside their patients' bodies. Prada and his team validated the idea first with researchers at Hospital Universitario Infanta Leonor, then with Spain's National Center for Cardiovascular Research. They realized they had found something worth building.
Aitheroscope works from a simple retinal photograph. The eye's blood vessels sit just behind the retina, visible and accessible without surgery, radiation, or contrast dye. The company's artificial intelligence model reads these images to identify signs of atherosclerosis before a patient develops symptoms. It is, in essence, a window into the vascular system. The technology was tested against carotid ultrasound—the clinical gold standard—in more than a thousand patients. The results were striking: roughly five percent false positives, three percent false negatives. The system caught patients who had fallen through the cracks of traditional risk assessment.
The distinction matters clinically. Traditional risk scales estimate the probability of a future heart attack. Aitheroscope detects plaque that is already present. When a doctor sees actual disease, not just statistical risk, they can act: recommend lifestyle changes, prescribe statins, establish structured follow-up. The patient moves from the low-risk category into active management.
In May 2026, Aitheroscope was named a finalist in the Fundación Mapfre Social Innovation Awards, competing against projects from Brazil, Mexico, and the United States. The company had already distinguished itself among nearly 470 applicants. But the real work was just beginning. Prada's team—ten people with deliberate diversity across clinical medicine, computer science, and business—had moved past validation into product consolidation and commercial expansion.
Nationally, Aitheroscope was already deployed across eighteen clinical centers in Spain, with negotiations underway in multiple regions to expand into both private and public healthcare systems. Internationally, the company had signed its first distribution agreement outside Spain, partnering with a cardiovascular technology specialist to bring the technology to Mexico and Chile, with plans to explore other Latin American markets. The agreement represented access to a region with significant potential through a partner with established networks.
The team was also extending the technology's reach. Because the retina reflects the overall state of the vascular system, the same approach could detect cerebrovascular disease and chronic kidney disease—conditions with enormous epidemiological impact. A complementary consumer-facing product was in development, aimed at cardiovascular risk awareness and democratizing access to screening. And in 2026, Horus ML was participating in the REACT study led by prominent cardiologists Valentín Fuster and Borja Ibáñez at the National Center for Cardiovascular Research—an external validation in one of the world's most significant cardiovascular cohorts.
The company's structure reflected its ambition. Carolina Espejo, the cardiologist who had first identified the problem, led clinical validation and hospital relationships. Santos Bringas, a computer vision specialist, directed product development as CTO. Yvonne Gala, a doctor in artificial intelligence with commercial experience, led strategy and partnerships. Prada himself held a doctorate in AI and served as both CEO and technical leader. None of this happened by accident. The cofounders shared a conviction that artificial intelligence should solve problems that mattered to human health. A family history of heart disease, a cardiologist's daily frustration with missed diagnoses, a technologist's refusal to optimize metrics instead of lives—these became the fuel. By 2026, Aitheroscope was moving from proof to scale, from Spain to the world, from detecting one disease to detecting many. The work of turning a real problem into a real solution was underway.
Notable Quotes
I knew from the start that I wanted to apply AI to something with real social impact, not just optimize business metrics. But I also knew it had to be in healthcare.— Jesús Prada Alonso, CEO of Horus ML
Subclinical atherosclerosis is massively underdiagnosed. Half of heart attacks happen in patients who were evaluated as low-risk, and the underdiagnosis is particularly severe in women.— Carolina Espejo, cardiologist and cofounder
The Hearth Conversation Another angle on the story
Why does a retinal image tell you anything about atherosclerosis? The eye seems far from the heart.
The retina isn't separate from the vascular system—it's part of it. The blood vessels in the back of your eye are the only ones doctors can see directly without cutting you open. They're a proxy for what's happening everywhere else in your body.
So you're reading the small vessels to infer what's happening in the large ones, like the carotid artery?
Exactly. And the validation proved it works. When we compared our AI's reading of retinal images to carotid ultrasound in over a thousand patients, we got it right about 97 percent of the time. That's better than the risk scales cardiologists have been using for decades.
But why does it matter that you're detecting disease that's already there, rather than predicting risk?
Because a doctor can't treat a probability. If I tell a patient they have a five percent chance of a heart attack in ten years, they might change their diet or they might not. But if I show them plaque in their arteries right now, they have a concrete reason to take a statin, to exercise, to be monitored. It's the difference between a warning and a diagnosis.
And the women—why were they being missed more than men?
The risk scales were built on populations that didn't represent women well. Women present differently, their symptoms are different, their risk factors interact differently. Half of all heart attacks in women happen in people classified as low-risk. We're catching those people now.
What happens next? You've validated it, you've deployed it in Spain. What's the real test?
The REACT study. It's the largest cardiovascular cohort in the world, led by some of the most respected cardiologists alive. If our technology holds up there, if it works in that population, then we know it's not just a Spanish success. We know it scales.