Currently submitted to: JMIR Cancer
Date Submitted: Jul 30, 2026
Open Peer Review Period: Jul 31, 2026 - Sep 25, 2026
(currently open for review)
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Evidence-Anchored Decision Support for Elective Neck Dissection in Oral Tongue Squamous Cell Carcinoma Using a Large Language Model and Multimodal Data
ABSTRACT
Background:
Indications for elective neck dissection (ND) in clinically node-negative (cN0) oral tongue squamous cell carcinoma (OTSCC) vary despite established risk factors, and existing AI tools often function as opaque, single-modality predictors that are difficult to audit at the bedside.
Objective:
We aimed to develop an explainable, evidence-anchored decision-support framework using a large language model with retrieval-augmented generation (LLM+RAG) that integrates routinely collected multimodal data into interpretable lymph-node (LN) risk estimates, with each factor's contribution linked to published evidence, and to evaluate its concordance with elective ND decisions and its discrimination for realized LN metastasis.
Methods:
In a single-institution retrospective cohort (n=85), literature-anchored, weighted prompts combined 11 modalities into 0-10 LN risk scores. Primary outcome was concordance with initial ND decisions; secondary outcome was discrimination for realized LN metastasis, with a prespecified T1/T2 subset.
Results:
Concordance with initial ND decisions was high (AUC 0.87; 95% CI, 0.80-0.95), whereas discrimination for realized metastasis was moderate (AUC 0.70; 95% CI, 0.59-0.82). In T1/T2 cases, scores did not separate metastasis outcomes (P=.11) but still mirrored ND decisions (AUC 0.79; 95% CI, 0.68-0.91).
Conclusions:
As a feasibility and interpretability assessment, this evidence-anchored LLM+RAG framework operating on routinely collected data produced interpretable LN risk estimates with case-level evidence citations and closely mirrored elective ND decisions, offering a transparent and auditable route to multimodal clinical decision support suitable for discussion at multidisciplinary tumor boards.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.