ImpactHealthcare5 MIN READ

Xlue Raises $35 Million Seed Round Led by Conviction for an AI Model That Predicts Cancer Risk From Health Records

Xlue, a Pittsburgh-based spinout from Carnegie Mellon University, has raised a $35 million seed round led by Conviction to develop AI that predicts a patient's cancer risk from routine electronic health records, according to funding data compiled in the 24 hours to November 19. Its CATCH-FM model is being tested at Kaohsiung Medical University Hospital in Taiwan for lung and liver cancer.

By Prathista Lazar · Author20 November 2026New
Xlue Raises $35 Million Seed Round Led by Conviction for an AI Model That Predicts Cancer Risk From Health Records

Many cancers are diagnosed too late, when treatment is more difficult and survival rates are lower. A young startup believes the warning signs may already be hidden in patients' medical histories.

Xlue, a Pittsburgh-based company spun out of Carnegie Mellon University in 2025, has raised a $35 million seed round led by Conviction, according to funding data compiled by AI Funding Tracker for the 24 hours to Thursday, November 19.

The round is unusually large for a seed-stage company, reflecting strong investor interest in AI applications in healthcare.

Predicting cancer before symptoms

Xlue's core technology is a foundation model called CATCH-FM, designed to predict a patient's risk of developing cancer based on their electronic health records. Health records contain information such as diagnoses, prescriptions, lab results and hospital visits over time. Patterns in this data can indicate an elevated risk of cancer, even before symptoms appear.

By identifying high-risk patients, the model could help doctors prioritise screening, such as scans or blood tests, for those most likely to benefit, leading to earlier diagnosis.

The model was trained on Taiwan's National Health Insurance Research Database, one of the world's largest health insurance databases, which covers almost the entire population of Taiwan.

In a study involving about 30,000 patients, 50% to 70% of those flagged by the model as high-risk were later diagnosed with cancer, according to information compiled by AI Funding Tracker.

Clinical trials

Xlue's model is being tested in a trial at Kaohsiung Medical University Hospital in Taiwan, focusing on lung and liver cancer. The company is also in talks with UPMC, the University of Pittsburgh Medical Center, one of the largest health systems in the US.

Clinical validation is essential for AI health tools, which must demonstrate that they are accurate, safe and useful in real-world settings before they can be widely adopted by hospitals and approved by regulators.

The founders

Xlue was founded by Carnegie Mellon researchers Ophir Frieder, Hao-Ren Yao and Chenyan Xiong, who serves as chief executive. Before the seed round, the company raised $1.5 million through a SAFE, a type of early-stage investment agreement.

Carnegie Mellon is one of the world's leading universities for computer science and AI research, and Pittsburgh has become a hub for AI, robotics and healthcare technology.

The investor

Conviction is a venture capital firm founded by Sarah Guo, a prominent AI investor and former partner at Greylock. The firm focuses on AI-native companies across infrastructure and applications.

The promise of AI in cancer care

AI is being applied across cancer care, from analysing medical images to detect tumours to predicting how patients will respond to treatment and accelerating drug discovery. Early detection is one of the most promising areas, because cancers caught early are often far more treatable.

Lung and liver cancers are among the deadliest forms of cancer, partly because they are often diagnosed at advanced stages. Lung cancer screening using low-dose CT scans is recommended for high-risk groups, such as heavy smokers, but many people who develop the disease do not fit these criteria. Better risk prediction could expand screening to people who would otherwise be missed.

The challenges

AI health prediction tools face significant challenges. Models trained on data from one population may not perform as well in others with different demographics, healthcare systems or record-keeping practices. False positives could lead to unnecessary tests and anxiety, while false negatives could provide false reassurance.

“The best cancer treatment is still an early diagnosis, and the clues may already be sitting in a patient's medical records.”
— TIGI Impact Desk

Privacy and data protection are also critical concerns when using patient records to train and run AI models. Regulators in the US, Europe and Asia are developing frameworks for AI medical devices, and companies must navigate complex approval processes.

cad85a3a-5fea-46c1-b971-05ac2a3657dc.png

What it means for India

India faces a growing cancer burden, with many cases diagnosed at late stages due to limited screening and awareness. The expansion of digital health records under the Ayushman Bharat Digital Mission could, over time, create opportunities to use AI for risk prediction and early detection. Indian healthtech startups are also developing AI tools for cancer screening, including for breast, oral and cervical cancer.

Why health records are valuable

Electronic health records capture a patient's interactions with the healthcare system over years, including diagnoses, procedures, medications and test results. Individually, many of these data points may seem unremarkable, but together they can reveal patterns that predict future disease.

AI models are particularly good at finding such patterns in large datasets, which is why researchers increasingly see health records as a powerful and underused resource for predictive medicine.

Taiwan's data advantage

Taiwan's single-payer National Health Insurance system, introduced in 1995, covers virtually all of its residents and has generated a comprehensive database of medical claims. Researchers worldwide have used the database for studies on disease patterns, treatments and outcomes.

The scale and completeness of the data make it well suited for training AI models, though models trained on Taiwanese data will need to be validated on other populations before being used widely elsewhere.

The investment trend

Investment in AI healthcare startups has grown rapidly, as investors see opportunities in diagnostics, drug discovery, clinical documentation and hospital operations. Large seed rounds for healthcare AI companies reflect confidence that AI can deliver meaningful improvements in patient outcomes and healthcare efficiency.

However, healthcare is a highly regulated industry with long sales cycles, and startups must demonstrate clinical value and navigate complex approval processes before achieving commercial scale.

Beyond lung and liver

While Xlue's initial trial focuses on lung and liver cancer, its foundation model approach could in principle be applied to other cancers and diseases that leave traces in health records. Foundation models, trained on large and diverse datasets, can often be adapted to new tasks with additional training, making them more flexible than models built for a single purpose.

If the approach proves successful, it could be expanded to predict risks of other serious conditions, such as heart disease, diabetes complications or kidney failure, potentially helping health systems shift from treating disease to preventing it. Achieving this would require extensive validation and close collaboration with clinicians.

What comes next

Xlue will use the funding to advance clinical trials, expand partnerships with hospitals and further develop its model. Its progress will be closely watched as a test of whether AI can make a meaningful difference in early cancer detection.

TagsXlueConvictionseed fundingcancer detectionearly diagnosisAI in healthcareelectronic health recordsfoundation modelCATCH-FMCarnegie Mellon UniversityChenyan XiongPittsburghTaiwanKaohsiung Medical UniversityUPMClung cancerliver cancerhealthtechimpactinvestors

Reader reviews

Sign in to rate and review this article.
Loading reviews…