New JMIR Cardio Section Seeks Research on Generative and Multimodal AI in Heart Care

Generative artificial intelligence is moving rapidly from research laboratories into the clinic, and cardiology is one of the specialties where the stakes are highest. Cardiovascular diseases remain the leading cause of mortality worldwide, and the promise of AI tools that can sharpen diagnosis, streamline clinical workflows, and widen access to specialist expertise has captured the attention of researchers, clinicians, and health systems alike. Yet the field has also learned a sobering lesson: a model that performs impressively on benchmark data may fail, or even cause harm, when deployed in the messy reality of everyday patient care. It is against this backdrop that JMIR Publications has announced a new section of its peer-reviewed journal JMIR Cardio, titled Generative and Multimodal AI in Digital Cardiovascular Medicine, and opened a call for submissions.

The new section, announced from Toronto on October 1, 2026, will publish peer-reviewed, time-sensitive research on the innovations, challenges, and open questions that sit at the intersection of artificial intelligence and cardiovascular medicine. JMIR Cardio is indexed in PubMed, PubMed Central, Sherpa Romeo, DOAJ, MEDLINE, CABI, and Scopus, and has met the editorial criteria for inclusion in the Web of Science Core Collection, placing it among the journals that clinical researchers typically watch for credible, citable work. By carving out a dedicated home for AI-focused cardiovascular research, the journal is signaling that the topic has matured from a novelty into a distinct field of inquiry that demands its own evidentiary standards.

The central theme the editors emphasize is a distinction that sounds obvious but is frequently ignored in the rush to publish: claims of benefit and evidence of benefit are not the same thing. Demonstrating technical performance alone, the announcement stresses, is insufficient to establish clinical value. A large language model that summarizes echocardiogram reports with high fidelity in a test environment still needs to prove that it saves clinician time without introducing errors, that it works across patient demographics, and that it does not quietly degrade the quality of care. Responsible implementation, the journal argues, requires rigorous validation, appropriate governance, and evaluation across diverse populations, health care systems, and real-world settings.

To that end, the section will prioritize research that evaluates the clinical value of AI in cardiovascular medicine rather than its algorithmic elegance. That includes evidence of effectiveness in real-world care, generalizability across populations and health care settings, impact on access and equity, feasibility of implementation and scaling, and costs. In practical terms, the editors are inviting studies that follow AI tools out of the laboratory and into hospitals, clinics, and patients’ homes, measuring what actually happens to workflows, outcomes, and health care spending when these systems go live. This orientation reflects a broader shift in the digital health literature, where the question is no longer simply whether a model can predict risk, but whether predicting risk changes anything that matters for patients.

The scope of the call is broad, spanning six thematic areas. The first covers clinical diagnosis, risk prediction, and decision support: generative AI, multimodal models, and large language models applied to cardiovascular diagnosis, risk stratification, treatment selection, clinical decision support, and personalized care planning. This is the territory where much of the current excitement concentrates, because cardiology generates exactly the kind of high-dimensional, heterogeneous data that modern AI architectures are built to digest. Electrocardiograms, echocardiograms, cardiac magnetic resonance images, laboratory panels, and longitudinal clinical notes each capture a different facet of cardiovascular health, and no single human clinician can integrate all of them at scale in real time.

That integration challenge defines the second thematic area: multimodal cardiovascular data fusion. The journal is particularly interested in clinically validated approaches that combine imaging, ECG and electrophysiological signals, electronic health records, laboratory data, patient-generated data, and mobile or wearable sensor streams. Multimodal models are technically demanding because different data types arrive in different formats, at different sampling rates, and with different noise characteristics, and aligning them meaningfully requires more than concatenating feature vectors. The emphasis on clinical validation is notable: the field has seen many multimodal architectures demonstrated on curated research datasets, but far fewer that have been tested prospectively on the unstructured, incomplete data that real clinics produce.

The third area addresses implementation and real-world cardiovascular care, including prospective evaluation, workflow integration, clinician-AI interaction, scalability, postdeployment performance, resource utilization, and the effectiveness and cost-effectiveness of AI-enabled care. This is arguably where the most consequential research gaps lie. Postdeployment monitoring, in particular, is a growing concern in the AI safety literature, because models can drift as populations, coding practices, and care patterns change over time. A risk prediction model validated in 2024 may silently lose calibration by 2028, and without systematic monitoring, neither developers nor clinicians would know. Studies that document these dynamics, or that test governance structures for catching them, would fill a genuine void.

The fourth thematic area turns toward patients themselves: conversational agents, natural language generation, and large language model-driven tools to enhance patient engagement, prevention, cardiac rehabilitation, remote care, and care pathway workflows. Crucially, the call also asks researchers to examine whether AI improves or inadvertently restricts access to cardiovascular expertise and services. This is a subtle but important framing. A chatbot that answers routine questions might free cardiologists to see complex cases, or it might become a gatekeeper that deflects patients who genuinely need human attention. Conversational agents in cardiac rehabilitation could extend supervised recovery programs to patients in remote areas, but only if the technology is usable by older adults, people with limited digital literacy, and those without reliable connectivity.

Equity, sex, and gender considerations form the fifth area, covering development and validation across diverse populations and health care settings, algorithm bias and digital exclusion, and sex- and gender-specific cardiovascular conditions, including maternal and perinatal cardiovascular health. The clinical rationale is strong: cardiovascular disease presents differently in women and men, conditions such as peripartum cardiomyopathy are inherently sex-specific, and training datasets that underrepresent women or minority populations can produce models that systematically underperform for those groups. Digital exclusion adds a further layer, since AI-enabled services delivered through smartphones or patient portals may be least accessible to the patients who carry the heaviest burden of cardiovascular disease.

The sixth area, evidence standards and responsible AI, may prove the most influential over time. The journal is seeking research that defines and evaluates the evidence required for clinical adoption of cardiovascular AI, including external validation, safety and failure analysis, human oversight, generalizability, postdeployment monitoring, data governance, and privacy and regulatory evaluation. In other words, the section is not only publishing AI studies but also inviting the field to debate what a sufficient evidence base for adoption should look like. That question is currently unresolved across medicine: regulators, professional societies, and payers are still working out how much and what kind of evidence an adaptive, continuously learning algorithm should be required to produce before and after it enters clinical use.

For researchers working at the intersection of cardiology and machine learning, the call represents an opportunity to publish work that will be judged against clinically meaningful criteria rather than leaderboard metrics alone. For clinicians and health system leaders, it promises a growing body of evidence about which AI tools actually deliver value under real conditions. And for patients, the ultimate stakeholders, the hope is that a research agenda built explicitly around effectiveness, equity, access, and safety will help ensure that the wave of generative and multimodal AI reaching cardiovascular medicine improves care rather than merely automating it. Submissions and further details are available through the JMIR Cardio website, and the publisher has positioned the new section as a venue for the time-sensitive, practice-relevant studies that this fast-moving field currently lacks a dedicated home for.

Subject of Research: Generative and multimodal artificial intelligence applications in digital cardiovascular medicine

Article Title: JMIR Publications’ JMIR Cardio invites submissions on Generative and Multimodal AI in Digital Cardiovascular Medicine

Article References: JMIR Publications’ JMIR Cardio invites submissions on Generative and Multimodal AI in Digital Cardiovascular Medicine. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: JMIR Cardio, generative AI, multimodal AI, cardiovascular medicine, large language models, clinical decision support, digital health, health equity, AI validation, patient-centered care, implementation science, wearable sensors