Short paper
A 6-8 page paper in Springer LNCS format describing the method, submitted via OpenReview, including Title, Authors, Abstract, Introduction, Methods, Results, and Conclusion.
NLP/CORE Unranked
HEADLINE 2026
HEADLINE: A Benchmark for Vision–Language Models in Head CT Reporting Challenge at MICCAI 2026
The HEADLINE 2026 Challenge is a MICCAI satellite event focused on automated radiology report generation from 3D head CT scans using Vision-Language Models. It provides a large-scale, clinically validated dataset of over 16,000 cases to develop AI systems that produce accurate, structured, and safe clinical reports, evaluated through a hybrid framework combining automated metrics and expert radiologist review.
Key deadlines
Verified September 9, 2026Full paper
July 26, 2026
AoE
Conference timeline
Submission and decisions
Full paperKey deadline
July 26, 2026 · AoE
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Research areas in scope
01
Challenge Task
Automated radiology report generationVision-Language Models for medical imaging3D head CT analysisClinical NLP in emergency neuroradiologyFinding identification (hemorrhage, infarction, fractures, tumors)Anatomical localization and clinical significance reporting
Policies worth checking twice
- Submissions must be provided as Docker containers executed on Grand Challenge infrastructure.
- Participants must comply with computational constraints: 16 or 32 GB DRAM, NVIDIA T4 (16 GB VRAM) or A10G (24 GB VRAM), 20 minutes per batch of up to 5 studies.
- Submissions must generate a report for every test case; failure results in a score of 0 for that case.
- Participants may use publicly available external datasets and pre-trained models, as long as they are accessible to all at challenge start.
- The dataset is not publicly available; access requires signing a Data Transfer and Usage Agreement (DUA) and is restricted to challenge use only.
- No commercial use of the dataset is permitted.
- Registration and method submissions are closed; only short papers via OpenReview are accepted.
- All accepted short papers will be publicly released on the MICCAI website one week before the conference; authors must file patents prior to this release.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026