ANTICIPATE: Artificial Intelligence to Improve Classification and Predict Malignant Transformation of Oral Epithelial Dysplasia

Overview

Oral epithelial dysplasia (OED) is a precancerous condition affecting the lining of the mouth. While most OED cases remain stable, a significant proportion progress to oral cancer—a disease with poor survival rates and rising incidence globally. Unfortunately, current diagnostic methods suffer from high variability and poor predictive accuracy.

ANTICIPATE is a multi-institutional research project funded by Cancer Research UK, aimed at transforming the diagnosis and risk stratification of OED using artificial intelligence (AI). Combining computational pathology, clinical expertise, and digital biomarkers, the project seeks to improve patient outcomes through earlier and more accurate detection of high-risk cases. See a list of selected publications below.

📢 Press release: Read about the launch of ANTICIPATE and its vision in our official press release.

As part of our commitment to translational research and real-world impact, the project culminated in a Stakeholder Engagement Day bringing together clinicians, researchers, and patient representatives. The event aimed to disseminate our findings, gather feedback, and spark discussions around future collaborations.

🔗 Read more in the TIA Centre news articleLink opens in a new window

Project Aims

Develop and validate AI models that better predict malignant transformation in OED
Identify useful biomarkers to help improve pathologist grading of OED

Method Overview

We collected over 500 histology slides from four centres: University of Sheffield, University of Birmingham, Queen’s University Belfast, and Piracicaba Dental School (Brazil).

Our models were trained on data from Sheffield and externally validated on the other cohorts to ensure generalisability. Throughout the project, we developed a range of tools for segmentation, grading, and outcome prediction.

In our final publication, we introduced ODYN — a transformer-based AI model for fully automated diagnosis and prognosis of oral epithelial dysplasia (OED). ODYN can:

Classify slides as normal or OED
Predict the risk of malignant transformation

It achieves accuracy comparable to pathologists, without any human intervention. Throughout this work, we have also identified the prognostic value of novel digital biomarkers, including:

Intra-epithelial lymphocytes (IELs): immune cells within the epithelium
Peri-epithelial lymphocytes (PELs): immune cells surrounding the epithelium.

Higher densities of these cells were found to be associated with increased cancer risk. See our publications below for more information.

Patient Impact

The ANTICIPATE project has supported research into OED, both developing prognostic models and finding new biomarkers to help inform clinicians. This will ultimately help patient outcomes by:

A step closer to providing consistent, objective diagnoses
Enabling earlier identification of high-risk lesions
Reducing over-treatment of low-risk cases
Improving prognostication with new digital biomarkers like IELs and PELs

These advances have the potential to improve clinical decision-making and support more personalised care pathways in oral cancer prevention.

Selected Publications

Literature Review
Mahmood H, et al. “Artificial Intelligence-based methods in head and neck cancer diagnosis: an overview” British Journal of Cancer, 2021.
👉 Read the paper
2- and 6-point Predive Models
Mahmood H, et al. “Prediction of malignant transformation and recurrence of oral epithelial dysplasia using architectural and cytological feature specific prognostic models.” Modern Pathology, 2022.
👉 Read the paper
HoVer-Net+
Shephard AJ, et al. “Simultaneous nuclear instance and layer segmentation in oral epithelial dysplasia.” ICCV Workshops, 2021.
👉 IEEE Xplore
PELs as Prognostic Marker
Bashir RMS, Shephard AJ, et al. “A digital score of peri-epithelial lymphocytic activity predicts malignant transformation in oral epithelial dysplasia.” Journal of Pathology, 2023.
👉 Read the paper
Nuclear/Architectural Features as Prognostic Markers
Mahmood H, Shephard AJ, et al. “Development and validation of a multivariable model for prediction of malignant transformation and recurrence of oral epithelial dysplasia.” British Journal of Cancer, 2023.
👉 Read the paper
OMTscore
Shephard AJ, Bashir RMS, et al. “A fully automated and explainable algorithm for predicting malignant transformation in oral epithelial dysplasia.” npj Precision Oncology, 2024.
👉 Read the paper
💻 Get the code
IEL Score
Shephard AJ, Mahmood H, et al. “A novel AI-based score for assessing the prognostic value of intra-epithelial lymphocytes in oral epithelial dysplasia.” British Journal of Cancer, 2024.
👉 Read the paper
💻 Get the code
🎮 Try the interactive demo
ODYN
Shephard AJ, Mahmood H, et al. “Development and validation of an artificial intelligence-based pipeline for predicting oral epithelial dysplasia malignant transformation” Communications Medicine, 2025.
👉 Read the paper
💻 Get the code
🎮 Try the interactive demo

Open-Source Code and Tools

As part of our commitment to open science and reproducibility, we have released code associated with several selected ANTICIPATE publications above.

All models are built using TIAToolbox, our open-source Python library for computational pathology, which now has over 200,000 downloads.

Team

University of Warwick, UK – Tissue Image Analytics (TIA) Centre

Prof. Nasir Rajpoot – Principal Investigator
Dr. Adam Shephard – Lead Research Fellow
Dr. Shan Raza – Faculty Member
Alumni: Dr. R.M.S. Saad Bashir

University of Sheffield, UK – School of Clinical Dentistry

Prof. Syed Ali Khurram – Principal Investigator
Dr. Hanya Mahmood – Lead Researcher
Alumni: Dr. Neda Azarmehr

Project Partners

University of Birmingham, UK

Queen’s University Belfast, UK

Piracicaba Dental School, São Paulo, Brazil

Dr. Anna Luiza Damaceno Araujo
Dr. Alan Roger Santos-Silva
Prof. Marcio Ajudarte Lopes
Prof. Pablo Agustin Vargas

Collaboration Opportunities

We are actively seeking new collaborators interested in further advancing this work. We welcome:

International centres willing to contribute annotated OED cases
Clinical partners interested in applying or testing our tools
Computational researchers interested in model interpretability or multimodal expansion (e.g. integrating radiology or genomics)

If you're interested in collaborating, please get in touch.

Contact

For academic or collaboration enquiries:

📧 n dot m dot rajpoot at warwick dot ac dot uk
📧 adam dot shephard at warwick dot ac dot uk

Highlights and News

🧠 New from the ANTICIPATE Study:

ODYN: Oral Dysplasia Network – Our latest AI tool predicts cancer risk directly from H&E biopsy slides using deep learning.

📅 Published May 2024 in Communications Medicine (Nature Portfolio)
🔗 Read the paper
💻 Explore the code
🎮 Try the interactive demo

IEL Score for Risk Stratification in OED – A novel AI-based score quantifying intraepithelial lymphocytes to stratify cancer risk in oral epithelial dysplasia.

📅 Published November 2024 in British Journal of Cancer
🔗 Read the paper
💻 Explore the code
🎮 Try the interactive demo

🗓️ Stakeholder Engagement Day

Disseminating Research, Driving Collaboration
Clinicians, researchers, and patient representatives joined us for a dedicated event to explore project outcomes and shape future directions.
🔗 Read moreLink opens in a new window (November 2024)