Call for Papers

MIDL 2027 · Porto, Portugal · 14-16 July

The Medical Imaging with Deep Learning (MIDL) Conference is an international forum for research at the intersection of deep learning, biomedical imaging, and healthcare. MIDL brings together researchers, clinicians, healthcare professionals, and industry to share methodological advances, rigorous validation studies, and insights into the responsible translation of artificial intelligence into clinical practice.

MIDL 2027 will be held in person in Porto, Portugal, with free online streaming to ensure global participation. The scientific programme will include invited talks, oral and poster presentations, and opportunities for exchange across disciplines.

We invite work spanning foundational research, technical innovation, and clinical translation.

We particularly welcome research that moves beyond performance on curated benchmarks and provides credible evidence of robustness, generalisation, clinical value, responsible development, or real-world impact.

Topics of Interest

Topics of interest include, but are not limited to:

  • Medical-image segmentation, detection, classification, registration, reconstruction, and acquisition
  • Computer-aided diagnosis, prognosis, treatment planning, and image-guided intervention
  • Self-supervised, semi-supervised, unsupervised, and data-efficient representation learning
  • Transfer learning, domain adaptation and generalisation, and learning with limited/noisy labels
  • Foundation models, vision-language models, and multimodal learning across imaging, text, omics, sensors, and clinical data
  • Generative AI, diffusion models, image synthesis, and synthetic data
  • Interactive tool-using and agentic AI for clinical and medical-imaging workflows
  • Uncertainty estimation, calibration, interpretability, explainability, and failure detection
  • Human+AI collaboration, clinical decision support, usability, and workflow integration
  • Federated, distributed, privacy-preserving, and secure learning
  • Safe, trustworthy, fair, transparent, and responsible AI
  • Data-centric AI, including dataset curation and annotation quality
  • Continual learning, post-deployment monitoring, model updating, and data-drift detection
  • Reproducibility, open science, robust benchmarking, and external, multicenter, or prospective validation across medical-imaging specialties

Full Papers

1. Main Track

The Main Track welcomes substantial methodological contributions to biomedical image analysis. Submissions may introduce new algorithms, architectures, learning strategies, evaluation methods, or theoretical insights. Authors should justify the clinical or biomedical relevance of the problem and evaluate their contribution against appropriate baselines.

Papers in this track may be up to 10 pages at submission (up to 12 pages after rebuttal), excluding references, acknowledgements, and appendices.

2. Special Track - Clinical Translation & Validation:

This track welcomes rigorous studies that evaluate, compare, or translate AI methods in clinically or operationally meaningful settings. Methodological novelty is not required when a submission offers an important scientific, clinical, or translational contribution. Relevant studies include:

  • External, multicentre, multi-site, temporal, or geographic evaluation
  • Prospective studies, silent deployment, and post-deployment monitoring
  • Reader studies and evaluations of human+AI teams
  • Clinical utility, patient outcome, workflow, usability, or health-economic studies
  • Reliability, robustness, fairness, and subgroup analyses
  • Evaluation under distribution shift, data drift, or changes in acquisition protocols and devices
  • Replication studies, and informative negative results

Submissions should clearly describe the study design, data provenance, reference standards, evaluation populations, statistical methods, and limitations. Authors should distinguish internal testing from genuinely external testing and report performance across clinically relevant subgroups where appropriate. Private data are welcome when their use is ethically justified but reproducibility through public data, shared code, executable models, or openly described protocols is strongly encouraged.

Papers in this track may be up to 14 pages at submission (up to 16 pages after rebuttal), excluding references, acknowledgements, and appendices.

Short Papers

Short papers may present promising early-stage work, novel ideas without extensive validation, position or perspective contributions, negative or replication results, or recent journal work of interest to the MIDL community.

Short papers are limited to 3 pages, excluding references. The main manuscript must be self-contained so, no appendices are allowed.

Please notice that the short paper track opens after the full-paper review cycle is completed (TBD).

Contacts

Inquiries to the program chairs can be addressed directly to [email protected].

Important DatesAuthor Instructions