New AI system could transform rabies testing

press release

Published: 12 October 2026

AI system could transform rabies testing around the world

Rabies virus detection in cell cultures could soon become faster and more accessible, thanks to a world-first AI system developed by researchers at the ��’s Institute for People-Centred AI (PAI) in collaboration with the Rabies and Viral Zoonoses team at the Animal and Plant Health Agency (APHA).

Credit: Getty Images

The breakthrough is designed to help improve surveillance of one of the world’s deadliest infectious diseases and deliver practical, laboratory-ready tools that can be adopted in endemic countries.

Despite being almost entirely preventable through vaccination, rabies causes around 59,000 human deaths every year. Current practices for diagnosis and measuring protective antibodies through vaccination still rely on specialists manually examining thousands of microscope images – which can be slow, labour-intensive and open to differences in interpretation.

In a DEFRA-funded study published in Nature Scientific Reports, the research team created the first AI-based workflow capable of identifying rabies virus infection in cell cultures and accurately measuring antibody concentration levels.

Using more than 2,300 fluorescence images containing over 36,000 annotated image sections, the team trained several deep learning models to recognise rabies virus infection within cells. The best-performing model correctly classified 99.6 per cent of samples and almost perfectly distinguished infected from uninfected cells, avoiding false alarms.

Rabies is one of the world’s oldest and deadliest infectious diseases, claiming tens of thousands of lives every year because timely diagnosis and surveillance remain challenging in many parts of the world. Once clinical signs appear, it is almost always fatal, so prevention and early diagnosis are essential.

"Our AI workflow provides laboratories with a faster and more consistent way to analyse samples while reducing the reliance on manual interpretation. Joan Manuel Amaya Cuesta, PhD Researcher, Centre for Vision, Speech and Signal Processing | Viral Zoonoses Research Scientist

The technology could help laboratories analyse samples more quickly and consistently, particularly in countries where rabies remains endemic and specialist expertise is limited. It could also reliably measure antibody levels used to assess vaccine protection and support international pet travel requirements.

The research reflects the ��’s Institute for People-Centred AI’s commitment to developing AI that addresses real-world challenges and delivers meaningful benefits for society. By combining advanced machine learning with the practical needs of diagnostic laboratories, the research team aims to create tools that support scientific expertise, improve consistency and help make reliable testing more accessible, particularly in regions where specialist resources are limited.

This research shows how AI can support scientists by carrying out complex image analysis quickly, consistently and at scale. While we've demonstrated this approach for rabies, the same framework could be applied to other infectious diseases that rely on similar laboratory tests, helping researchers and diagnostic laboratories make better use of AI in the future. Professor Miroslaw Bober, Professor of Video Processing and Machine Intelligence

The results establish a framework that can be adapted for the detection, analysis, and surveillance of other pathogens, supporting broader applications of AI across infectious disease research and diagnostics.

We hope this innovation will help strengthen rabies surveillance and, ultimately, support global efforts to eliminate human deaths from this disease. This technology is now being further developed for use to detect the virus in infected brain tissue which we hope will transform international rabies diagnostic laboratories. Dr Lorraine M. McElhinney, Head of the National Reference Laboratory for Human and Animal Rabies at the APHA

The breakthrough is translating cutting-edge AI architectures into a practical workflow for rabies testing. We evaluated convolutional neural network (CNN) and transformer-based architectures, including MambaOut and Swin Transformer V2, with MambaOut achieving the strongest performance at 99.6% accuracy in classifying culture wells in our evaluation. Our workflow connects infection detection in microscope images with antibody quantification, matching expert assessments of antibody levels on previously unseen test images. This marks an exciting step towards more automated, consistent, and scalable laboratory analysis, with the potential to strengthen rabies surveillance and vaccine assessment. Dr Syed Sameed Husain, Lecturer in People-Centred AI

###

Notes to editors

Professor Miroslaw Bober is available for interview; please contact mediarelations@surrey.ac.uk to arrange.
The full paper can be found here:
Paper authors: , , , &