Agentic AI for Personalised and Clinical Task-Based PET Image Reconstruction

About the Project

Positron emission tomography (PET) is a powerful tool for studying the brain, but conventional image reconstruction methods use fixed algorithms and assumptions that cannot adapt to each patient’s unique data or clinical question.

This project introduces a new generation of intelligent PET image reconstruction using agentic artificial intelligence (AI): AI systems that can reason, adapt, and interact like human experts. Instead of applying one-size-fits-all settings, the AI will adjust the reconstruction process in real time, optimising parameters and priors based on the specific diagnostic task (for example, epilepsy or dementia) and even for different regions of the brain.

In the first year, the project will benchmark state-of-the-art methods, including advanced diffusion-based priors, to establish the best achievable image quality. In the second year, multiple cooperating AI agents will be developed to guide and refine reconstruction based on clinical intent, image content, and diagnostic feedback. By the third year, the system will evaluate its own reliability and uncertainty, producing confidence maps that show how trustworthy each part of the image is. This combination of agentic and generative AI aims to deliver clearer, more informative PET images tailored to each patient, supporting more accurate diagnoses and better clinical outcomes.

Funding Notes

Full funding for 3.5 years, covering tuition fees, a stipend, and a research support allowance as indicated below

Tuition Fees: Covered for Home students ONLY.

Stipend: £23,805 (2026/27 UKRI rate), paid monthly and tax-free.

Research Training Support Grant (RTSG): Up to £10,500 over 3.5 years for research costs and conference attendance.

Informal enquiries to supervisors are welcome, but all candidates must follow the application process outlined on the Multiscale Medical Imaging website (View Website) and submit an application via King’s College London by the stated deadline to be considered.