Answers are grounded in this paper and cite their sources.
Demonstrates accurate classification and prognosis in head and neck small round blue cell neoplasms, suggesting improved diagnostic strategies.
Aims: Head and neck small round blue cell neoplasms (SRBCNs) comprise biologically distinct tumours with overlapping histopathological features that often complicate diagnosis. This study aimed to develop and validate a multimodal artificial intelligence framework for diagnostic classification and prognostic stratification of SRBCNs. Methods: A multicentre cohort of 675 cases was categorised into haematolymphoid (n = 463), neuroectodermal and neural crest-derived (n = 84), mesenchymal (n = 75), and melanocytic (n = 53) groups. Whole-slide images were analysed using quantitative nuclear morphometry, deep learning architectures, and foundation models (UNI and Virchow). Clinicopathological and morphometric variables were incorporated into an XGBoost classifier and survival model. External validation was performed using an independent cohort. Results: Significant differences in nuclear morphometry were identified among tumour groups (all p < 0.0001). The XGBoost classifier achieved an accuracy of 98.0% and a macro-AUC of 0.9996. Foundation models demonstrated the highest performance, with Virchow achieving an accuracy of 92.4% and AUROC of 0.981 internally, and 89.8% and 0.972 during external validation. Age and nuclear morphometric features were the most influential classification variables. Nuclear area remained independently associated with overall survival (HR = 1.02, p = 0.012), while the XGBoost survival model achieved a C-index of 0.768. Conclusions: Multimodal artificial intelligence enables accurate classification and prognostic stratification of head and neck SRBCNs. The integration of histological, morphometric, and clinicopathological data provides objective biomarkers that may support diagnostic decision-making and precision oncology.
No takes yet. Share an insight, caveat, or question.
Souza et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: