MICCAI Workshop on Multimodal Learning with Medical Tabular Data
In conjunction with the 29th International Conference on Medical Image Computing and Computer Assisted Intervention, Strasbourg Convention Center, Room Adenauer (U), France
Tabular clinical data—including electronic health records (EHRs), laboratory measurements, omics data, and other clinical variables—remain underutilized in medical imaging research despite their widespread availability in biobanks and clinical repositories. Their heterogeneous structure, variable coding schemes, and substantial missingness make them difficult to model and integrate with imaging and other modalities.
MultiTab focuses on advancing methodological approaches that address these challenges, positioning tabular and semi-structured medical data as a core component that can enhance medical imaging analysis. We aim to gather researchers developing AI methods for learning from diverse tabular data types and for combining them with image data and other complementary modalities.
Workshop Program
16:00 - 16:05
Opening remarks
16:05 - 16:50
Keynote: Beyond Echocardiography: Toward Clinical Diagnosis Through Multimodal Fusion
Olivier Bernard, INSA Lyon - CREATIS
16:50 - 17:20
Oral Session
17:20 - 17:55
Poster Session
17:55 - 18:00
Final Remarks & Awards
Accepted Papers
View accepted papers, reviews, and official workshop submissions on the MultiTab OpenReview page.
We welcome contributions across the following topics:
Novel architectures and techniques for integrating imaging with tabular data (e.g., clinical, omics, EHRs) and other modalities (e.g., text, time-series, signals).
Generalization, robustness, and trustworthiness in models combining tabular and imaging data with focus on scalability, interpretability, and uncertainty quantification.
Adaptation, fine-tuning, and evaluation of foundation models to enhance fusion and representation learning across tabular and imaging data.
Cross-modal interpretability, fairness, bias mitigation, and compliance in models leveraging both imaging and structured tabular data.
Standards, datasets, and evaluation metrics specifically designed for multimodal research involving imaging and tabular data.
Techniques for missing data, noise, multi-source integration, domain shifts, and longitudinal multimodal analysis with temporal alignment.
Submission Portal
OpenReview
Submit your paper through OpenReview
Submission Guidelines
Follow MICCAI Standards
We follow the MICCAI main conference submission guidelines for paper format, length, and quality standards to ensure consistency across all MICCAI 2026 events.
Supplementary material: Supplementary material submissions are not supported for this workshop. If additional resources are necessary for reproducibility (e.g., detailed attribute lists, more implementation details), authors are encouraged to provide them through an associated public code repository. All information essential for reviewing and understanding the work must be included in the main paper.
Double-blind review
8 pages + references
LNCS format
01/07 07/07/2026
Paper Submission Deadline
Submit your full paper through OpenReview
July 31, 2026
Acceptance Notifications
Authors notified of acceptance/rejection decisions
August 7, 2026
Camera-Ready Version Deadline
Final accepted papers due
September 27, 2026
Workshop Date
2-hour workshop at MICCAI 2026 in Room Adenauer (U), Strasbourg, France
Talk Title
Beyond Echocardiography: Toward Clinical Diagnosis Through Multimodal Fusion
Abstract
Echocardiography is the first-line modality for establishing cardiac diagnoses. However, developing reliable automated tools on top of this modality is often constrained by data scarcity, interobserver variability, and domain shifts across centers and vendors. Recent foundation models illustrate a shift toward multimodal integration, combining ultrasound sequences with clinical reports for automatic diagnosis and report generation at scale. Yet these data-intensive approaches still lack visual evidence clinicians could inspect, explicit uncertainty modeling, and interpretability - leaving predictions difficult to trust or explain.
This talk presents an alternative strategy that leverages clinical knowledge to reduce reliance on massive datasets while reinforcing explainability. By extracting interpretable time series from ultrasound sequences - through robust segmentation, uncertainty estimation, and motion tracking - we bridge raw imaging data with clinically meaningful descriptors such as global and regional strain. We then show how these time series can be fused with patient tabular data (EHR) to characterize arterial hypertension, using transformer-based tokenizers and an ordinal classification scheme, before introducing a more advanced asymmetric fusion framework that decouples shared and modality-specific information to improve both performance and interpretability.
Finally, we discuss current limitations - restricted cohort size, simplified data representations, and the open challenge of more advanced explainability - and outline next steps toward clinically actionable, trustworthy multimodal diagnostic tools.
Speaker Bio
Prof. Bernard received his Electrical Engineering degree and Ph.D. from the University of Lyon (INSA), France, in 2003 and 2006, respectively. He was a Postdoctoral Fellow with the Biomedical Imaging Group at the Federal Polytechnic Institute of Lausanne, EPFL, Switzerland in 2007.
He is currently Professor at the University of Lyon (INSA) and the Deputy Director of the CREATIS laboratory in France. He is Senior member of the Institut Universitaire de France (IUF). He is also the head of the MYRIAD research team, which specializes in medical image analysis, simulation, and modeling.
His current research interests focus on image analysis through deep learning techniques, with applications in cardiovascular imaging, blood flow imaging, and population representation. From a methodological point of view, Prof. Bernard's recent research focuses on heterogeneous data integration, domain adaptation, and uncertainty estimation using AI-based methods.
Prof. Bernard is currently Associate Editor (AE) of the IEEE Transactions on Ultrasonics journal, AE of the European Heart Journal (Cardiovascular Imaging), and guest AE of the Transactions on Medical Imaging.
