SRS92 Patey Prize Entrant - The AIRSPACE Project: redefining prognostic stratification in advanced laryngeal cancer through radiomics discovery and international validation—an international multicentre observational cohort study
Abstract Objectives Prognostic stratification in advanced laryngeal cancer remains imprecise and no validated biomarkers exist to guide treatment. The AIRSPACE Project is the first large-scale programme to investigate whether radiomic features from routine pre-treatment CT scans could reveal prognostic markers not captured by standard clinicopathological assessment. In this sub-study, we examined whether combining CT-derived radiomic features with clinical factors could improve prediction of overall survival and validated the model internationally. Design Retrospective–prospective multicohort observational study. Setting Northern England with independent validation across five international centres via The Cancer Imaging Archive (TCIA). Participants 374 patients with stage T3–T4 laryngeal squamous cell carcinoma (246 discovery, 128 validation). Main outcome measures Overall survival. Results The clinical model showed poor discrimination (internal C-index 0.60, AUC 0.61; external C-index 0.61, AUC 0.61). The radiomics model improved performance (internal C-index 0.65, AUC 0.69; external C-index 0.66, AUC 0.69). The combined model performed best, achieving strong discrimination in both internal (C-index 0.73, AUC 0.78) and external validation (C-index 0.70, AUC 0.77). Patients classified as high risk by the Rad-score had more than double the risk of death (HR 2.47, 95%c.i. 1.23–4.92). Conclusions The AIRSPACE Project provides the first large-scale evidence that a combined radiomic–clinical model improves prognostic stratification in advanced laryngeal cancer, validated across international cohorts. Radiomics unlocks underutilised information already present in routine CT scans, and its integration with clinical factors offers a strong foundation for future adoption in patient care and trial design.
Authors
- Khaled Kallas
- Amar Rajgor
- A. Mill
- David W. Hamilton (ORCID: https://orcid.org/0000-0002-9653-6453)
- Bogusław Obara (ORCID: https://orcid.org/0000-0003-4084-7778)
- Andrew McQueen
- Steven Rushton
- Eric Aboagye
Institutions
- Newcastle upon Tyne Hospital (GB)
- NIHR Newcastle Biomedical Research Centre (GB)
- Imperial College London (GB)
- Newcastle University (GB)
Publication Details
- Journal
- British journal of surgery
- Published
- 2026-03-01
- DOI
- https://doi.org/10.1093/bjs/znag018.012
- Primary Topic
- Radiomics and Machine Learning in Medical Imaging
- Type
- article
- Field-Weighted Citation Impact
- 0.00