Nine months is the difference between a death sentence and a fighting chance
El cáncer de páncreas ha sido durante décadas una sentencia casi inmediata: silencioso, tardío en manifestarse y veloz en su desenlace. Una startup española llamada Sycai ha desarrollado un algoritmo de inteligencia artificial capaz de detectar esta enfermedad aproximadamente nueve meses antes que un radiólogo humano, un margen que, en una dolencia donde la supervivencia media se mide en estaciones, puede transformar una condena en una oportunidad. La tecnología no reemplaza al médico, sino que amplía su mirada, integrándose en los sistemas hospitalarios como un segundo par de ojos que nunca se cansa.
- El cáncer de páncreas mata en cuatro o cinco meses de media, y casi siempre se descubre demasiado tarde porque no da señales hasta que ya ha avanzado.
- Los radiólogos modernos enfrentan escáneres de doscientas imágenes en apenas quince minutos, una carga que hace casi inevitable que ciertas lesiones pasen desapercibidas.
- El algoritmo de Sycai, entrenado en miles de imágenes, identifica marcadores biológicos, reconstruye lesiones en tres dimensiones y alerta al especialista antes de que el tumor sea visible para el ojo humano.
- Tras cuatro años de desarrollo, patentes y certificación regulatoria, la herramienta ya opera en hospitales europeos desde 2025, con plena anonimización de datos y sin que la información salga del entorno hospitalario.
- Sycai apunta ahora al cáncer de hígado y a herramientas de guía quirúrgica que permitan operar con precisión en casos donde el tumor invade vasos sanguíneos.
El cáncer de páncreas llega sin avisar. La mayoría de los pacientes lo descubren por casualidad, durante una consulta por otro motivo, cuando el escáner revela la mala noticia. Para entonces, casi siempre es demasiado tarde. La supervivencia media es de cuatro a cinco meses. Incluso los radiólogos más experimentados pueden pasarlo por alto: el páncreas se esconde en lo profundo del abdomen, fragmentado en decenas de cortes de imagen, fácil de ignorar cuando se busca otra cosa.
Sycai, una startup española, ha construido una inteligencia artificial para encontrar lo que el ojo humano no ve. Su software detecta el cáncer de páncreas aproximadamente nueve meses antes que un especialista, según los ensayos que le valieron la certificación regulatoria. En una enfermedad donde cada mes importa, nueve meses es la diferencia entre no tener opciones y tener alguna.
Sara Toledano, CEO y cofundadora, explica el problema con claridad: los escáneres modernos generan hasta doscientas imágenes por exploración, y un radiólogo dispone de apenas quince minutos para revisarlas todas. Se busca lo que se espera encontrar. El resto queda invisible. El algoritmo de Sycai trabaja de otro modo: entrenado en miles de casos, reconoce la forma, la textura y la posición de las lesiones, las reconstruye en tres dimensiones y lanza la alerta. Toledano lo describe como un radiólogo con vidas enteras de experiencia acumulada, que nunca se cansa ni pierde el foco.
El software se integra directamente en los sistemas hospitalarios. El radiólogo ve el escáner junto al análisis de la IA, confirma el diagnóstico y tiene siempre la última palabra. Los datos permanecen anonimizados y nunca salen del entorno del hospital. Tras cuatro años de desarrollo y certificación, la herramienta lleva operativa en hospitales europeos desde 2025.
El páncreas es solo el comienzo. Sycai trabaja ya en la detección del cáncer de hígado y en herramientas de guía quirúrgica que permitan mapear con precisión dónde puede operar un cirujano cuando el tumor invade vasos sanguíneos. En 2026, España diagnosticará cerca de 10.400 nuevos casos de cáncer de páncreas. Para esos pacientes, nueve meses no es una estadística. Es tiempo.
Pancreatic cancer arrives quietly. Most patients discover it by accident—they visit a doctor for something else entirely, and the scan reveals the bad news. By then, it is almost always too late. The disease kills fast. Median survival is four to five months. Even radiologists, trained to read thousands of images, can miss it. The pancreas sits deep in the abdomen, fragmented across dozens of scan slices, easy to overlook when you are searching for something else.
A Spanish startup called Sycai has built artificial intelligence to find what radiologists miss. The software detects pancreatic cancer roughly nine months before a human specialist would, according to trials that earned it regulatory certification. In a disease where every month matters—where the entire window of survival is measured in seasons—nine months is the difference between a death sentence and a fighting chance.
Sara Toledano, the company's CEO and co-founder, explains the problem plainly. Modern CT scanners produce far more data than they used to. A decade ago, a scan might yield twenty slices. Today it yields two hundred, each one sharper than before. This is progress, in theory. In practice, it means radiologists face an impossible workload. A specialist has roughly fifteen minutes to examine all those images. They look for what they expect to find. Everything else—the tumor hiding in plain sight—goes unseen.
Sycai's algorithm works differently. It has been trained on thousands of images to recognize the biological markers of pancreatic cancer: the shape of a lesion, its texture, the way it sits among surrounding tissue. The software measures it, renders it in three dimensions, flags the warning signs. Toledano describes it this way: imagine a radiologist with not five or ten or twenty years of experience, but lifetimes of it. A radiologist who never tires, never loses focus, never runs out of time. That is what the algorithm does.
The software integrates directly into hospital systems. When a radiologist opens their work queue, they see the latest scans alongside Sycai's analysis. If a lesion is present, they are alerted. The specialist confirms the diagnosis, reviews the characterization, compares it to previous scans to track how the cancer has evolved. The doctor always has the final word. The AI is a second set of eyes, not a replacement.
Radiologists have embraced it. They work with technology constantly. Similar AI tools already exist for breast and lung cancer. There is no mystique, no resistance. The real barrier is regulatory. Data must be anonymized. Patient information never leaves the hospital. It is stored on the hospital's own servers or cloud infrastructure, never passing through Sycai's systems. The company spent four years developing, patenting, and certifying the pancreatic cancer tool before launching it across European hospitals in 2025.
The pancreas is only the beginning. Sycai is now developing detection for liver cancer, another disease that hides until it is advanced. Beyond that, the company envisions something more ambitious: surgical guidance. When cancer invades blood vessels, surgery becomes impossible—the risk of hemorrhage is too great. Sycai wants to map exactly where a surgeon can safely cut, turning a guessing game into precision work.
In 2026, Spain expects roughly 10,400 new cases of pancreatic cancer. It will rank far below colorectal, breast, and lung cancers in frequency. But it will kill faster and more completely than almost any other malignancy. For those patients, nine months is not a statistic. It is the difference between no options and some options. It is time.
Notable Quotes
The pancreas has a very specific position in the body that looks quite poor on CT scans, shown in pieces and in different positions, and it's a fairly aggressive cancer.— Sara Toledano, CEO of Sycai
The radiologist has 15 minutes to see all those images, so many times they go exclusively to see what they're looking for; we want what isn't being looked for to be found as well.— Sara Toledano, CEO of Sycai
The Hearth Conversation Another angle on the story
Why is pancreatic cancer so hard to catch in the first place?
It sits deep in the abdomen, fragmented across scan slices, and it doesn't announce itself. Patients feel fine until they don't. By then it's stage four.
So the radiologist is looking at two hundred images in fifteen minutes and missing something that's hiding anyway?
Exactly. They're searching for what they expect to find. The algorithm doesn't have that bias. It looks at everything.
But doesn't a doctor need to confirm what the AI finds? Doesn't that just add another step?
It does, but now the doctor is looking for something specific. The AI has already narrowed the field. It's like the difference between searching a dark room and searching a dark room with a flashlight.
Nine months is a long time in a disease where people have four or five months to live. How does that change things?
It doubles the survival window. It means surgery might be possible. It means treatment options exist. It means the patient gets to make choices instead of just accepting the inevitable.
What happens next for Sycai?
Liver cancer next. Then surgical guidance—helping surgeons know exactly where they can cut safely. The goal is to turn these invisible killers into manageable problems.