Multimodal Biomedical Data & Interpretable AI: Synergies and Applications

Dear Colleagues,

Biomedical research and healthcare are entering a phase in which data availability is no longer the limiting factor, but rather the ability to integrate heterogeneous information into reliable, actionable knowledge. Large-scale initiatives and routine clinical practice are now generating multimodal data at an unprecedented level of depth and breadth. This includes electronic health records (EHRs), medical imaging, digital pathology, genomics and other 'omics' layers, biosignals (ECG/EEG), wearable streams and, increasingly, biomedical text and knowledge resources. At the same time, the rapid adoption of AI in clinical and translational settings has exposed a central tension: while complex models can achieve strong predictive performance, their opacity, vulnerability to shifts in distribution, and uncertain generalisability can limit trust, safety, and regulatory acceptance. This necessitates a decisive shift towards multimodal biomedical learning coupled with interpretability, where performance improvements are not pursued in isolation but rather in alignment with transparency, causal reasoning, and clinical utility.

This research area is particularly timely because global trends are converging. Health systems are under pressure to improve efficiency and equity while facing rising burden due to chronic disease and ageing populations. Meanwhile, precision medicine demands patient-level stratification and more nuanced decision support. As such, federated and privacy-preserving analytics are becoming essential as data governance tightens and cross-institutional collaboration increases. These dynamics present significant opportunities: multimodal fusion can reduce uncertainty by triangulating evidence across data types, while interpretable AI can generate mechanistic hypotheses, facilitate clinician oversight and accelerate discovery by linking phenotypes to molecular pathways. However, significant challenges remain, including modality-specific biases and missing data, inconsistent coding and annotation, label noise, confounding factors, domain shifts across hospitals and devices, and the difficulty of producing faithful, clinically meaningful and robust explanations. Progress requires methodological innovation, rigorous evaluation and reproducible pipelines that connect algorithmic advances to constraints of real-world deployment.

The aim of this Special Issue is to present and disseminate the most recent advances in the integration of multimodal biomedical data and interpretable AI, emphasising synergies between modalities and demonstrable impact in research and practice. We welcome contributions that address methods, benchmarks and applications that fuse heterogeneous biomedical data (e.g. EHR, imaging, pathology, omics, time series, text and knowledge resources) while ensuring interpretability, calibration, robustness and clinical or biological plausibility.

Topics of interest for publication include, but are not limited to, the following:

• Multimodal learning architectures for biomedical data (late/early fusion, cross-attention, multimodal transformers)
• Data fusion under missing, noisy, or partially observed modalities; weak supervision and label harmonization
• Interpretable AI (XAI): feature attribution, concept-based explanations, counterfactuals, rule extraction, and uncertainty-aware explanations
• Causal inference and causal representation learning for clinical prediction and biomarker discovery
• Integration of EHR with imaging, pathology, omics, and biosignals (ECG/EEG/wearables) for risk stratification and prognosis
• Knowledge graphs, ontologies, and retrieval-augmented models for grounded biomedical reasoning
• Calibration, uncertainty quantification, out-of-distribution detection, and robustness to dataset shift
• Fairness, bias auditing, and subgroup performance analysis in multimodal clinical AI
• Privacy-preserving and federated multimodal learning; secure computation and governance-aware analytics
• Benchmark datasets, standardized evaluation protocols, and clinically meaningful endpoints
• Prospective/retrospective validation, deployment studies, and human–AI collaboration in decision support
• Reproducible workflows, model reporting standards, and best practices for translation to regulated environments

Dr. Alexandre G. De Brevern
Guest Editor

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the Special Issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 250 words) can be sent to the Editorial Office for assessment.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-anonymized peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. BioMedInformatics is an international peer-reviewed open access semimonthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 1200 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.