Abstract
Head and neck cancer (HNC) management requires accu- rate tumor delineation, disease staging, and prognosis to support clini- cal decision-making. The HEad and neCK TumOR (HECKTOR) 2026 challenge addresses these tasks by introducing a unified benchmark for primary tumor (GTVp) and metastatic lymph node (GTVn) segmenta- tion, TN staging, and recurrence-free survival (RFS) prediction from multimodal PET/CT imaging and clinical data. We propose a mod- ular framework integrating established methods for segmentation, TN staging, and RFS prediction. For segmentation, we adopt the Adaptive Multi-modality Segmentation-to-Survival model (AdaMSS), which learns an adaptive fusion strategy for combining complementary anatomical in- formation from CT and metabolic information from PET. To enable fully automatic inference, the original ground-truth-based cropping strategy of AdaMSS is replaced with a scan-specific anatomical-prior localization procedure. TN staging uses independent CatBoost classifiers trained on compact handcrafted radiomic, anatomical, and clinical features. Finally, RFS prediction uses a Random Survival Forest integrating segmentation- derived features, clinical variables, and probability distributions from the TN staging models. On the HECKTOR 2026 validation set, the frame- work achieved a weighted score of 0.6630, Mean Dice of 0.6388, T-stage and N-stage balanced accuracies of 0.6155 and 0.5379, respectively, and an RFS C-index of 0.7536. On the held-out test set, it achieved a weighted score of 0.5920, Mean Dice of 0.6386, T-stage and N-stage balanced accu- racies of 0.5237 and 0.6032, respectively, and an RFS C-index of 0.5879, placing the team among the top five challenge submissions. These re- sults support the feasibility of integrating established methods within a unified modular framework.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/HECKTOR_003.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=IeaB6Z0Kf8
BibTex
@InProceedings{MehYas_Deep_MICCAISAT2026,
author = { Mehmood, Yasar AND Hu, Christian Anchele AND Zaidi, Maryam Fatima},
title = { { Deep Learning-Based Segmentation and Feature-Based Clinical Prediction for Head and Neck Cancer } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17271},
month = {pending},
page = {pending}
}
