HERD: Towards integration of advanced MRI and multi-omic biomarkers for early relapse prediction in high-risk head and neck cancer-a prospective study
No Thumbnail Available
Authors
Mithra S.
Amato F.
Gomes C.
Ahmed A.
van Schelt A.S.
Cherukara M.
Rowley M.
Coolen T.
ReisFerreira M.
Sawyer E.
Contact
Check for full-text access
Issue Date
2026
Type
Conference Proceedings
Language
Keywords
Alternative Title
Abstract
Topic Imaging, radiomics and artificial intelligence Keywords HPV-negative, multi-omics, predictive modeling Purpose/Objective Head and Neck Squamous Cell Carcinoma (HNSCC) is the seventh most common cancer worldwide, with over 12,000 new cases annually in the UK. Patients with locally advanced, HPV-negative disease experience poor outcomes despite radical therapy (approximately half relapse within two years)<sup>1,2</sup>. Current surveillance strategies are largely symptom-driven and lack sensitivity for early, subclinical recurrence. The HERD (Head and Neck Early Recurrence Detection) study aims to develop and validate a multimodal, biologically informed model integrating advanced MRI, radiomics, and multi-omic biomarkers to improve early relapse detection and stratification in this high-risk cohort. Material/Methods HERD is a CRUK Early Detection-funded, multi-centre prospective study (opened 2023 - completion 2027) recruiting up to 200 patients with radically treatable stage-III/IV, HPV-negative HNSCC. Participants undergo longitudinal sampling over two years, including multi-parametric MRI, biofluids (blood, saliva, urine, stool), and tumour tissue collection (Figure 1). HERD tests three hypotheses: advanced MRI metrics of hypoxia and stromal architecture reflect tumour microenvironment (TME) states predictive of relapse; that genomic and immune signatures stratify recurrence risk; and microbiome composition influences tumour behaviour and treatment response (Figure 2). Five analytical platforms underpin HERD. Advanced MRI - including MR Elastography, Quantitative Susceptibility Mapping, and DCE-MRI - yields quantitative stiffness, oxygenation, and perfusion maps, analysed radiomically to identify spatial TME biomarkers. Immune profiling uses multiparameter Flow Cytometry, CyToF, and high-plex immunofluorescence to define systemic and tissue-resident phenotypes. Genomic profiling of cfDNA uses targeted sequencing to detect genetic mutations validated against matched tumour tissue. Patient-derived organoids model treatment response ex-vivo, while microbiome sequencing (16S rRNA) of oral, urine, and stool samples characterise microbial diversity. Data integration occurs by harmonising multimodal datasets into "meta-covariates" and analysing via AI-driven modelling (following Bayesian frameworks) to generate personalised risk scores. Results 32 patients are enrolled across three London centres (Guy's & St Thomas', UCLH, and Barts), with recruitment expanding nationally. Standardised protocols for imaging and biological sample collection have been implemented, enabling harmonised data capture. Early feasibility suggests successful acquisition of quantitative MRI and radiomics features alongside high-quality biofluid data, demonstrating strong potential for multi-domain data integration as HERD progresses. Conclusion HERD represents one of the first large-scale, prospective studies to combine advanced MRI, radiomics, AI, and multi-omic profiling in high-risk HNSCC. Through multimodal data integration, the study seeks to elucidate mechanisms of relapse, establish a framework for precision, risk-adapted surveillance, therefore enabling earlier and improved interventional access in this this challenging population. Copyright © 2026 Elsevier Ireland Ltd
Description
Citation
Publisher
License
Journal
Radiotherapy and oncology : journal of the European Society for Therapeutic Radiology and Oncology
Volume
Issue
ICHNO 2026.