HECKTOR 2026

Deadlines
AI/CORE Unranked

HECKTOR 2026

Fifth Edition of Head and Neck Tumor Lesion Segmentation, Staging and Prognosis using Multimodal Data

1790852400000Strasbourg, FranceOfficial conference site Site reachable

HECKTOR 2026 is a challenge held in conjunction with MICCAI 2026 that advances machine learning solutions for head and neck cancer by unifying tumor segmentation, radiological TN staging, and recurrence-free survival prediction into a single end-to-end clinical pipeline. It leverages a large, multi-institutional dataset of multimodal FDG PET, CT, and clinical data to promote the development of robust, generalizable, and clinically translatable AI models.

Paper fit

Contribution paths

A strong submission should clearly identify its contribution and evaluate it appropriately.

Docker-based algorithm submission

Participants submit Docker containers that process multimodal imaging and clinical data to perform segmentation, TN staging, and prognosis prediction in a single pipeline, with evaluation across three phases: Sanity Check, Validation, and Testing.

Conference paper

Teams must submit a 6–12 page paper in LNCS format via OpenReview to be eligible for official ranking; the paper must describe the primary approach, results, and methodology used in the challenge.

Research areas in scope

01

Challenge Tasks

Primary tumor (GTVp) segmentationLymph node (GTVn) segmentationRadiological TN stagingRecurrence-free survival prediction

Policies worth checking twice

  • All testing phase submissions must be accompanied by a paper to be eligible for ranking.
  • Docker containers must not exceed 10 GB in size.
  • Inference time per patient must not exceed 25 minutes.
  • Submissions must run on an ml.g4dn.2xlarge instance with 1× NVIDIA T4 GPU, 8 vCPUs, and up to 32 GiB RAM.
  • Participants may submit up to 3 Docker containers for the Validation Phase and must select 1 for the final Testing Phase.
  • Papers must be in PDF format, using Springer’s LNCS templates, and include the team name at the end of the abstract.
  • Teams must cite the two official HECKTOR dataset papers in their submissions.

Official sources

Compiled from the official call for papers. The organizers’ pages remain authoritative.

Last verified September 9, 2026