Getting a PET scan can be stressful, costly, and time-consuming. From fasting and radioactive injections to long waiting times and post-scan isolation, the process has remained largely unchanged for decades. But a new Oxford spinout, RADiCAIT, is working to revolutionize this experience, using artificial intelligence to transform ordinary CT scans into synthetic PET scans, offering a faster, more affordable alternative.
The startup, which recently came out of stealth with $1.7 million in pre-seed funding, aims to make advanced diagnostic imaging accessible to hospitals and clinics that lack the expensive infrastructure PET scans require. RADiCAIT is one of the Top 20 finalists at TechCrunch Disrupt 2025 and has opened a $5 million seed round to fund clinical trials and regulatory approvals.
Turning Simple CT Scans into AI-Generated PET Scans
At the core of RADiCAIT’s innovation is a generative deep neural network developed at the University of Oxford in 2021 by co-founder and Chief Medical Information Officer Regent Lee. The model learns by comparing large datasets of CT and PET scans, identifying patterns that link anatomical structures from CTs to the physiological information found in PETs.
Chief Technologist Sina Shahandeh describes the process as translating “distinct physical phenomena,” effectively teaching the AI to read the body in multiple ways — structure and function — simultaneously. Multiple models then collaborate to produce the final synthetic image that doctors can analyze. The approach is similar in concept to Google DeepMind’s AlphaFold, which transformed biology by predicting protein structures from genetic data.
CEO Sean Walsh says their technology can “mathematically prove” that these synthetic PET images are statistically equivalent to real ones. Early trials show that radiologists and oncologists make the same quality of clinical decisions when presented with AI-generated PETs as they do with chemical PETs.
Breaking Barriers in Accessibility and Cost
Traditional PET scans are limited because the radioactive tracers used must be produced in specialized cyclotrons and used within hours, confining most scanners to large hospitals in major cities. Rural and regional patients often face long travel times and high costs to access these services. RADiCAIT’s solution could eliminate those bottlenecks by leveraging existing CT scan machines, which are far more common and affordable.
“PET is one of the most constrained and costly medical imaging modalities,” said Walsh. “What we’ve done is replace that with the most accessible — CT — without sacrificing accuracy.”
Clinical Trials and the Road Ahead
RADiCAIT has already partnered with Mass General Brigham and UCSF Health for pilot studies focusing on lung cancer diagnostics. The startup is now preparing for FDA-approved clinical trials, with plans to expand into colorectal and lymphoma imaging next. The company’s broader goal is to prove commercial viability and show that AI can significantly cut healthcare costs while maintaining clinical precision.
While RADiCAIT doesn’t plan to replace PET scans entirely — especially in therapeutic contexts like radioligand therapy — it aims to take over the diagnostic and monitoring roles, freeing up traditional PET scanners for treatment use.
A Future Beyond Radiology
According to Shahandeh, the implications of RADiCAIT’s technology go far beyond oncology. “Our approach, using AI to discover hidden relationships between different scientific domains, can be applied across radiology, materials science, and even chemistry,” he said.
By bridging the gap between affordability and accuracy, RADiCAIT could redefine how doctors detect and monitor disease, offering a glimpse into a future where medical imaging is not only smarter but also available to everyone, regardless of geography or income.
