Benchmarking Deep Learning Models for Laryngeal Cancer Staging Using the LaryngealCT Dataset

2025-10-13

AI Summary

Longstanding limitations in laryngeal cancer imaging analysis stem from the absence of standardized, publicly available CT datasets, hindering reproducible deep learning research. To address this, we introduce LaryngealCT—the first open-source benchmark dataset for laryngeal T-staging—comprising 1,029 contrast-enhanced CT scans. We propose a weakly supervised region-of-interest (ROI) extraction method and integrate 3D Grad-CAM for interpretable model analysis, revealing attention patterns around critical anatomical structures, particularly the laryngeal cartilages. In binary classification tasks—early (Tis–T2) versus late (T3–T4) staging, and T4 versus non-T4—we achieve state-of-the-art AUCs of 0.881 and 0.892 using a 3D CNN and ResNet18, respectively. All models undergo rigorous validation by clinical experts. The dataset, source code, and trained models are fully open-sourced to foster transparency and reproducibility in laryngeal cancer AI research.

Technology Category

• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Web Mining and Content Analysis: Large pretrained models with web data
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Laryngeal cancer imaging research lacks standardised datasets to enable reproducible deep learning (DL) model development. We present LaryngealCT, a curated benchmark of 1,029 computed tomography (CT) scans aggregated from six collections from The Cancer Imaging Archive (TCIA). Uniform 1 mm isotropic volumes of interest encompassing the larynx were extracted using a weakly supervised parameter search framework validated by clinical experts. 3D DL architectures (3D CNN, ResNet18,50,101, DenseNet121) were benchmarked on (i) early (Tis,T1,T2) vs. advanced (T3,T4) and (ii) T4 vs. non-T4 classification tasks. 3D CNN (AUC-0.881, F1-macro-0.821) and ResNet18 (AUC-0.892, F1-macro-0.646) respectively outperformed the other models in the two tasks. Model explainability assessed using 3D GradCAMs with thyroid cartilage overlays revealed greater peri-cartilage attention in non-T4 cases and focal activations in T4 predictions. Through open-source data, pretrained models, and integrated explainability tools, LaryngealCT offers a reproducible foundation for AI-driven research to support clinical decisions in laryngeal oncology.

Problem

• Standardizing datasets for reproducible laryngeal cancer deep learning models
• Benchmarking 3D architectures for early versus advanced cancer staging
• Providing explainable AI tools for clinical decision support in oncology

Innovation

• Used weakly supervised parameter search for CT scan extraction
• Benchmarked 3D deep learning models for cancer staging
• Integrated explainability tools with 3D GradCAMs for predictions