Abstract
HECKTOR2026 comprises segmentation,radiological T and N staging, and recurrence-free survival (RFS) prediction from pretreat- ment FDG-PET/CT and clinical data. We present JOTAC-HN, a modular cascade that shares predicted tumour masks across these tasks. A CT- anchored cranial crop precedes an adaptive three- or five-model residual- encoder nnU-Net ensemble. Its masks define anatomical, metabolic, spatial and radiomic features for separate CatBoost staging models and an Individual Coefficient Approximation for Risk Estimation (ICARE) model. Downstream training used masks from fold models trained with- out the corresponding patient. The cohort comprised 782 patients from eight centres, with centre-grouped evaluation for staging and prognosis. Public validation scores were 0.6697 mean Dice, 0.6120 T-stage balanced accuracy, 0.7511 N-stage balanced accuracy, 0.5786 RFS C-index and an overall weighted score of 0.6374. On the hidden test set of about 400 cases from three unseen centres, the later refitted submission achieved mean Dice 0.6447, balanced accuracies of 0.5267 for T staging and 0.5917 for N staging, an RFS C-index of 0.6266 and a weighted score of 0.6075. Team: JOTAC.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/HECKTOR_015.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=rGnHdvNoNO
BibTex
@InProceedings{NegAni_JOTACHN_MICCAISAT2026,
author = { Negi, Aniket AND Tully, Aditya AND Gupta, Ankur},
title = { { JOTAC-HN: Joint Outcome and TN Assessment Cascade for Head and Neck Cancers } },
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}
}