Oxford spinout RADiCAIT leverages artificial intelligence to lower costs and expand access to diagnostic imaging — see it in action at TechCrunch Disrupt 2025

Anyone who has undergone a PET scan knows it can be quite a challenge. While these scans are crucial for doctors to identify cancer and monitor its progression, the procedure itself can be very inconvenient for patients. 

The process begins with fasting for four to six hours before arriving at the hospital — and if you live in a rural area without access to a PET scanner, the situation is even more complicated. Upon arrival, you’re given an injection of radioactive material, then required to wait about an hour for it to circulate through your system. Afterward, you must remain as still as possible inside the PET scanner for half an hour while the radiologists capture the images. Once finished, you need to avoid close contact with children, pregnant women, and the elderly for up to 12 hours, since you remain partially radioactive.  

Another challenge is that PET scanners are mostly found in large cities, since the radioactive tracers they use must be made in nearby cyclotrons — small nuclear reactors — and used within a few hours, making them hard to access for patients in less populated regions. 

But what if artificial intelligence could transform CT scans — which are much more widely available and less expensive — into PET scans? That’s the concept behind RADiCAIT, a company spun out of Oxford that recently emerged from stealth mode with $1.7 million in pre-seed funding. The Boston-based firm, a Top 20 finalist in Startup Battlefield at TechCrunch Disrupt 2025, has just launched a $5 million fundraising round to further its clinical research.  

“Our approach was to take the most restricted, complicated, and expensive imaging method in radiology and replace it with the most accessible, straightforward, and cost-effective one — CT,” explained Sean Walsh, CEO of RADiCAIT, in an interview with TechCrunch. 

The core of RADiCAIT’s technology is its foundational model — a generative deep neural network developed in 2021 at the University of Oxford by a team led by co-founder and chief medical information officer Regent Lee.  

The model is trained by comparing CT and PET images, learning to map and recognize the relationships between them. According to RADiCAIT’s chief technologist, Sina Shahandeh, the system essentially links “different physical phenomena” by converting anatomical details into physiological data. The model is then guided to focus on particular scan features, such as certain tissues or irregularities. This targeted training is repeated with numerous examples, enabling the model to distinguish which patterns are medically significant. 

The final image that doctors review is produced by integrating several models working in tandem. Shahandeh likens this method to Google DeepMind’s AlphaFold, the AI that transformed protein structure prediction: Both technologies translate one form of biological data into another.  

Walsh asserts that RADiCAIT’s team can mathematically demonstrate that their AI-generated PET images are statistically equivalent to traditional chemical PET scans.  

“That’s what our studies confirm,” he said, “that the quality of decisions made by doctors, radiologists, or oncologists is the same whether they use a chemical PET or our AI-generated PET.” 

RADiCAIT isn’t aiming to eliminate PET scans entirely, especially in therapeutic contexts like radioligand therapy for cancer treatment. However, for diagnosis, staging, and monitoring, their technology could potentially make PET scans unnecessary.   

“Because this system is so limited, there simply aren’t enough resources to satisfy the demand for both diagnostics and theragnostics,” Walsh explained, referencing a medical strategy that merges diagnostic imaging (such as PET scans) with targeted therapies for disease treatment. “Our goal is to address the diagnostic demand, while PET scanners can focus on theragnostic applications.” 

RADiCAIT has already started clinical pilot programs for lung cancer screening with major healthcare providers like Mass General Brigham and UCSF Health. The company is now seeking FDA approval for a clinical trial — a more complex and costly step that is the main reason for its $5 million seed round. Once approved, the next phase will be commercial pilots to prove the product’s market potential. RADiCAIT also plans to follow the same process — clinical pilots, clinical trials, and commercial pilots — for colorectal cancer and lymphoma. 

Shahandeh noted that RADiCAIT’s method of using AI to deliver reliable results without the need for difficult or expensive tests has “broad potential.” 

“We’re working on expanding this approach throughout radiology,” Shahandeh continued. “You can expect to see similar breakthroughs connecting fields from materials science to biology, chemistry, and physics wherever there are hidden patterns in nature to uncover.” 

If you’re interested in learning more about RADiCAIT join us at Disrupt, October 27 to 29 in San Francisco. Learn more here.